v0.6.5 — Queen Phase Separation 👑

Highlights

Queen Phase Separation

The queen agent's responsibilities are now cleanly separated into three phases, each with its own scoped system prompt and toolset.

• Building phase — Full coding tools for writing and editing agent code, designing graphs, and constructing worker agents
• Staging phase — Agent is loaded but not running; queen can inspect, configure, and launch the worker
• Running phase — Worker is executing; queen monitors, controls, and can stop or restart
• Dynamic prompt swapping — System prompt and available tools change on each phase transition, reducing prompt size and improving instruction adherence
• QueenPhaseState — New state machine managing phase lifecycle, tool availability, and phase-change event notifications
• Smaller model support — Phase-scoped prompts significantly improve instruction following on smaller models

Queen Thinking Hook

• Thinking visibility — New thinking hook surfaces queen reasoning during agent construction
• Max token control — Configurable token limits for thinking output

Telegram Tool Expansion

Expanded Telegram integration with message management, media, and chat info operations. (#5403)

---

What's New

Core Framework

• Queen prompt optimization — Condensed building phase prompts, removed unused prompts, reorganized tool documentation
• Escalation system — New escalation tool for worker-to-queen communication during execution
• Progressive disclosure — Runtime data revealed progressively to avoid information overload
• Dynamic system prompts — System prompts loaded dynamically into context
• Validation improvements — GCU and validation added to initialize_agent_package; agent.json validated before parsing (#5846)
• Judge improvements — Judge can now wait for queen input; skip-judge logic improved
• BOM-safe JSON loading — Handles byte-order marks in JSON files (#5901)
• LLM logger on by default — LLM debug logging enabled by default
• Config error logging — Parse errors now logged for better debugging (#4955)
• Provider key detection — Improved indirect variable expansion for API key detection (#5504)
• Minimax provider — Added Minimax provider mapping and stream fallback

Frontend & Server

• Duplicate session fix — Prevent duplicate session creation when starting from home
• CLI validation — Validate --output path before agent execution (#5838)

Refactoring

• Remove old TUI CodingAgent — Deprecated TUI-specific coding agent removed
• Remove unused builder functions — Cleaned up legacy builder code and output cleaner
• Remove old skills — All unused agent skills removed
• Rename queen mode → phase — Consistent "phase" terminology throughout codebase
• Reorganize tools — Coder tools and agent initialization tools restructured

---

Bug Fixes

• Fix back-to-back compaction edge case
• Fix turn signal edge case
• Fix output key that terminated the queen
• Fix duplicated session calls
• Fix Cannot write to closing transport error
• Fix missing GCU prompts and instructions
• Fix thinking hook max token limits
• Fix MCP tools and templates loading
• Fix logger schema mismatch
• Fix legacy agent.json loading error handling
• Fix Aden client import duplication after rebase
• Skip POSIX permission tests on Windows (#5842, #5847)

Documentation

• Add README for brevo, csv, runtime_logs, account_info tools (#5602)
• Remove old agent skill references
• Restructure docs layout

---

Community Contributors

• Jack (@jackthepunished) — Telegram tool expansion with message management, media, and chat info (#5403)
• VasuBansal7576 — Minimax provider integration

---

Upgrading

git pull origin main
uv sync

# For the web frontend
cd core/frontend
npm install
npm run build

or, simply run

./quickstart.sh (linux/mac)
.\quickstart.ps1 (windows)

v0.6.4 — 300+ Tools 🍯

Highlights

60+ New Tool Integrations

The largest integration expansion in Hive's history — over 60 new tool integrations (300+ individual tools) landed in a single release, many contributed by the community. Agents can now connect to CRMs, databases, cloud services, messaging platforms, and more out of the box.

CRM & Sales — Salesforce, Zoho CRM, Attio, Pipedrive, Lusha
Project Management — Jira, Asana, Linear, Trello, Notion
Developer Tools — GitLab, Docker Hub, Vercel, Databricks, Redis, Supabase, HuggingFace, Pinecone, Terraform Cloud
Communication — Twilio, Zoom, Microsoft Graph (Outlook, Teams, OneDrive), Pushover, Tines, Brevo
Data & Analytics — Snowflake, Amazon Redshift, Azure SQL, Apache Kafka, Power BI, Google Sheets, Google Search Console, MongoDB, Airtable, MSSQL
Finance — Plaid, QuickBooks, Yahoo Finance
Cloud & Storage — AWS S3, Cloudinary
Media & Content — YouTube Data API, YouTube Transcript, Reddit, Confluence, Obsidian, Langfuse
Automation — n8n, Calendly, Apify, PagerDuty
Other — Twitter/X, Greenhouse, DuckDuckGo, SAP S/4HANA, Shopify

Each integration includes credential specs, health checks, and unit tests.

Hashline Edit Tool

A new cross-tool hashing anchor system for precise file editing.

• Hash-based line anchoring — Unique hashes per line for stable references across tool calls
• Grep search integration — grep_search with large file skip reporting
• Coder tools support — Hive Coder agent tools now support hashline editing
• CRLF handling — Prevents double-conversion in hashline edits

Windows Platform Support

Comprehensive Windows compatibility improvements across the entire stack.

• Filesystem support — Windows file system paths and operations (#5677)
• UTF-8 encoding — Systematic enforcement across tools and core to fix charmap decode errors
• Browser auto-open — Fixed quickstart browser launch on Windows
• os.fchmod guard — Safely handles missing POSIX calls on Windows

Session Management & Continuity

• Session resume — Conversations resume from where they were left off
• Session loading — All sessions load properly without errors
• Pause/stop fix — Pipeline pause now correctly uses stop_worker like queen

---

What's New

Core Framework

• Verified/unverified tool tiers — Tool loading split into verified and unverified tiers for clearer trust boundaries. (core/framework/tools/)
• Google OAuth unification — Single credential flow for all Google services (Calendar, Sheets, Search Console). (tools/)
• Google Scopes expansion — Extended OAuth scopes for broader Google API access. (#5764)
• Remove old session tools — Deprecated get_agent_session_state and get_agent_session_memory removed. (#5828, #5829)
• Remove hardcoded Anthropic logics — LLM provider logic is now fully configurable.
• Health check system — Simplified health check architecture with LLM key validation.
• LLM provider key update — Easy mechanism to update LLM provider keys.
• GCU enabled by default — Graph context updates now on by default.
• Recommended models update — Updated default model recommendations.
• Mac keychain for Claude Code — Utilize macOS keychain for Claude Code subscription credentials.

Frontend & Server

• hive open command — New CLI command to launch the browser-based interface directly.
• Permanent top bar — Fixed persistent top navigation bar.
• Quickstart improvements — Auto-install Node 20, Windows alignment, streamlined setup flow.

New Agent Templates

Template	Description
Twitter News Agent	Automated Twitter/X news monitoring and aggregation
Local Business Agent	Local business discovery and information
Meeting Scheduler Agent	Calendar-aware meeting scheduling with Google Calendar

---

Bug Fixes

• Fix permanent top bar display issue
• Fix pause in pipeline to use stop_worker like queen (#5816)
• Fix CRLF double-conversion in hashline edit and add large file skip reporting
• Fix Windows compat — guard os.fchmod and remove deleted LLM_CREDENTIALS import
• Fix UTF-8 encoding across tools and core to resolve Windows charmap decode errors (#5786)
• Fix browser auto-open after quickstart on Windows (#5797)
• Fix stdin conflicts in quickstart flow
• Fix quickstart build failing (#5834)
• Fix session loading and resume continuity (#5815)
• Fix event routes unhandled error
• Remove duplicate _execute_subagent method in EventLoopNode (#5785)
• Remove old get_agent_session_state and get_agent_session_memory tools
• Fix tool tests and credential wiring for community integrations
• Remove hardcoded Anthropic logics from LLM provider

Documentation

• Update roadmap to reflect completed features (refs #4780)
• Update AGENTS.md with latest conventions
• Documentation updates for hive open command
• Update recommended models documentation

---

Community Contributors

A huge thank you to all community contributors who made this release possible — especially the massive wave of tool integrations!

• Aaryann Chandola (@Antiarin) — Hashline edit tool with cross-tool hashing anchors (#5294)
• Fernando Mano (@fermano) — Windows filesystem support and testing (#5814)
• Kostas Foskolakis (@kostasuser01gr) — Roadmap documentation updates (#5774)
• code-Miracle49 (@code-Miracle49) — Remove duplicate _execute_subagent method (#5785)
• Shaurya Singh (@Waryjustice) — Fix browser auto-open on Windows (#5797)
• Omar Shareef (@osb910) — UTF-8 encoding enforcement for Windows charmap errors (#5786)
• Levin (@levxn) — Session management and conversation continuity (#5815)
• Amdev-5 (@Amdev-5) — X (Twitter) integration with DM support; Lusha MCP integration (#3814)
• Vasu Bansal (@VasuBansal7576) — AWS S3, SAP S/4HANA, and Plaid integrations
• Navya Bijoy (@navyabijoy) — Databricks MCP tool integration
• KRYSTALM7 (@KRYSTALM7) — Pushover push notification integration
• Utkarsh Singh (@Hustple) — Brevo transactional email and SMS integration
• Sayeed Rizwan (@iamSayeedR) — Asana integration for project management
• y0sif (@y0sif) — Attio CRM and Linear project management integrations
• mubarakar95 (@mubarakar95) — Apify integration
• Shivam Shahi (@shivamshahi07) — Zoho CRM integration
• mishrapravin114 (@mishrapravin114) — Calendly and Airtable integrations
• kubrakaradirek (@kubrakaradirek) — MSSQL tool with schema discovery
• Manas Dutta (@manasdutta04) — Reddit integration with 18 MCP tools
• Sapna Vishnoi (@Sapnavishnoi) — Amazon Redshift integration
• Aneesh (@rooney011) — YouTube transcript integration
• GastonAQS (@GastonAQS) — Trello integration
• Sashank Thapa — X (Twitter) integration tool
• P Gokul Sree Chandra (@GokulAIx) — YouTube Data API integration
• Gordon Ng (@DrNogNog) — QuickBooks integration
• hh3538962 (@hh3538962) — Power BI integration

---

Upgrading

git pull origin main
uv sync

# For the web frontend
cd core/frontend
npm install
npm run build

or, simply run

./quickstart.sh (linux/mac)
.\quickstart.ps1 (windows)

v0.6.3

Highlights

Queen Mode Separation

The queen agent now operates in three distinct modes with dynamic tool provisioning.

• Building mode -- Queen has full coding tools for editing and creating agents
• Staging mode -- Worker agent is loaded and ready, queen gains lifecycle tools (start, stop, pause, inject messages)
• Running mode -- Worker is executing, queen switches to monitoring and control tools (get status, stop worker, inject messages)
• Frontend mode sync -- Mode tag displayed on queen messages with visual styling per mode; frontend can trigger mode transitions

Interactive Question Widget

A new structured question UI for agent-to-human interactions.

• Option-based questions -- Agents can present up to 3 predefined options plus an auto-appended "Other" free-text option
• Keyboard shortcuts -- Number keys (1-4) to select options, Enter to submit
• ask_user tool upgrade -- Options support added to the ask_user tool for structured user input collection
• Queen context endpoint -- New /queen-context API for queuing non-intrusive messages to the queen between iterations

Email Reply Agent Template

A new sample agent demonstrating email automation workflows.

• Email filtering -- Intake node for user criteria collection
• Search and reply -- Automated unreplied email search with draft composition
• Confirmation gate -- Recipient approval required before sending
• Continuous conversation mode -- Multi-node conversation threading support

---

What's New

Core Framework

• Queen mode state machine -- Three-stage mode switching (building/staging/running) with dynamic tool provisioning and event-driven transitions. (core/framework/tools/queen_lifecycle_tools.py)
• Execution quality tracking -- Granular metrics: clean/degraded/failed execution quality, retry counts, node visit tracking, and partial failure detection. (core/framework/graph/executor.py)
• Continuous conversation mode -- Phase-aware conversation compaction with phase_id, transition markers, and client input tagging. (core/framework/graph/conversation.py)
• Consolidated MCP tool discovery -- Refactored discover_mcp_tools and list_agent_tools into a single flow. (core/framework/graph/event_loop_node.py)
• Agent collaboration guidelines -- New AGENTS.md for multi-agent workspace conventions. (AGENTS.md)
• uv instructions -- Added uv usage guidance to agent system prompts. (core/framework/agents/hive_coder/nodes/__init__.py)

Frontend & Server

• QuestionWidget component -- Interactive option-based question UI with keyboard navigation and free-text fallback. (core/frontend/src/components/QuestionWidget.tsx)
• Queen mode display -- Mode tag on queen messages with per-mode styling and frontend mode switching controls. (core/frontend/src/pages/workspace.tsx, core/frontend/src/components/ChatPanel.tsx)
• Queen context endpoint -- POST /api/sessions/{id}/queen-context for queuing external events without triggering immediate LLM response. (core/framework/server/routes_execution.py)
• Queen status on session start -- Returns queen bee mode state when starting a session. (core/framework/server/routes_sessions.py)
• Credential modal fix -- Hide delete button for Aden-managed credentials. (core/frontend/src/components/CredentialsModal.tsx)

New Agent Template

Template	Description
Email Reply Agent	Email filtering, search, reply composition, and confirmation-gated sending with continuous conversation mode

---

Bug Fixes

• Fix pause/stop to cancel all running tasks across all graphs in TUI (#5234)
• Fix queen auto-block from overwriting pending worker questions
• Fix queen returning on empty stream
• Fix conversation compaction logics for multi-phase execution
• Fix LLM conversation cleansing for clean context windows
• Fix MCP server cwd resolution from repo root instead of agent path
• Fix credential store sharing between server and runner to avoid redundant syncs
• Fix hide delete button for Aden-managed credentials
• Fix load new session from home page with proper session ID generation
• Fix execution recovery for interrupted sessions
• Fix initial state condition for queen mode transitions
• Fix agent generation guidelines for hive coder

Documentation

• Remove TUI references from primary README
• Add running screenshot and update coding agent instructions
• Sync all i18n READMEs with primary README (es, hi, ja, ko, pt, ru, zh-CN)
• Reorder documentation sections
• Update latest features in README

---

Community Contributors

• Aaryann Chandola (@Antiarin) -- Fix pause/stop to cancel all running tasks across all graphs (#5234)

---

Upgrading

git pull origin main
uv sync

# For the web frontend
cd core/frontend
npm install
npm run build

v0.6.2

Highlights

Separate Queen and Worker Input

A significant UX redesign for multi-agent workflows.

• Dedicated input channels -- The main chat input is now permanently connected to the queen agent, eliminating routing ambiguity
• Inline worker reply boxes -- When a worker requests user input, a dedicated reply box appears inline in the chat thread
• Clearer multi-agent interactions -- Users always know which agent they're communicating with

Subagent Framework

A major architectural addition enabling hierarchical agent orchestration with queen-worker communication patterns.

• Queen-worker communication -- Parent agents can spawn and coordinate child subagents with bidirectional messaging
• Subagent lifecycle tools -- New tools for spawning, monitoring, and managing subagent execution
• Progressive feedback -- SubagentJudge provides incremental feedback during long-running subagent tasks
• Worker status reporting -- Consolidated status updates from workers to parent agents
• Subagent logging -- Organized log structure with subagent logs stored in dedicated node folders

GCU (General Computer Use)

New browser automation capabilities built into the framework as a first-class node type.

• Browser control tools -- Complete toolkit for navigation, interaction, inspection, and tab management
• Interactive element highlighting -- Visual highlighting of actionable elements on web pages
• Snapshot tools -- Page state capture for debugging and verification
• Health checks -- Browser session validation at startup
• GCU node type -- Built-in browser tools automatically available for GCU-enabled nodes
• Quickstart integration -- GCU option added to the quickstart menu

Security Hardening

Multiple security improvements for safer agent execution.

• Agent path validation -- Restricts agent loading to allowed directories only
• OAuth token permissions -- Enforces 0600 permissions on OAuth token files
• Preload validation -- Enhanced pre-start validation for GCU subagents and credentials

Frontend & UX Improvements

Significant updates to the web workspace experience.

• Colorful tool pills -- Visual distinction with color-coded tool call indicators
• Agent building animation -- Animated feedback during agent construction
• Subagents in node panel -- Subagent activity displayed in the node detail panel
• Scheduler countdown -- Frontend display of agent idling timer
• Separate worker/queen input -- Distinct input handling for orchestrated agents

---

What's New

Core Framework

• Subagent framework -- Queen-worker communication with bidirectional messaging and lifecycle management. (core/framework/tools/queen_lifecycle_tools.py, core/framework/graph/event_loop_node.py)
• GCU node type -- Browser automation as a built-in node capability with tool highlighting. (core/framework/graph/gcu.py, tools/src/gcu/)
• Agent idling detection -- Runtime detection of idle agents with configurable timeouts. (core/framework/runtime/agent_runtime.py)
• Preload validation -- Enhanced validation pipeline for subagents and credentials before agent start. (core/framework/runner/preload_validation.py)
• LLM debug logger -- Human-friendly logging for LLM calls and tool invocations. (core/framework/runtime/llm_debug_logger.py)
• Gemini 3.1 Pro support -- Added support for Gemini 3.1 Pro model. (core/framework/llm/litellm.py)

Frontend & Server

• Tool pills v2 -- Colorful, per-call tool status indicators with turn tracking. (core/frontend/src/components/ChatPanel.tsx)
• Agent building animation -- Visual feedback during agent graph construction. (core/frontend/src/pages/workspace.tsx)
• Node panel subagents -- Subagent activity visible in node detail panel. (core/frontend/src/components/NodeDetailPanel.tsx)
• Scheduler countdown -- Timer display for agent idling detection. (core/frontend/src/pages/workspace.tsx)
• SSE event handling -- Improved event bus to prevent critical event queue blocking. (core/framework/runtime/event_bus.py)

Tools & Integrations

• GCU browser tools -- Full browser automation suite:
  • Navigation: goto_url, go_back, go_forward, refresh_page
  • Interactions: click_element, type_text, scroll, drag_and_drop
  • Inspection: get_page_content, get_element_text, take_screenshot
  • Tabs: new_tab, switch_tab, close_tab, list_tabs
  • Advanced: execute_javascript, wait_for_element, extract_data
• File operation tools -- Upgraded file operations with better error handling. (tools/src/aden_tools/file_ops.py)

Documentation

• Runtime initialization docs -- Comprehensive documentation for runtime startup flow. (docs/runtime_initialization.md)
• Architecture updates -- Expanded architecture documentation. (docs/architecture/README.md)
• GCU guide -- Reference guide for browser automation in hive coder. (core/framework/agents/hive_coder/reference/gcu_guide.md)

---

Bug Fixes

• Fix queen agent flakiness in orchestration scenarios
• Fix unused /hive-credentials prompts in validation flow
• Fix auto-block grace period check ordering before _await_user_input
• Fix GCU error log throwing
• Fix conversation ordering in multi-turn interactions
• Fix tool pill turn ID tracking
• Fix agent loading issues
• Fix browser snapshot tool excessive calls
• Fix subagent tool call overflow margin inheritance
• Fix subagent spillover handling
• Fix SSE critical event queue blocking event bus
• Fix timer countdown update in scheduler
• Fix Windows compatibility in quickstart scripts
• Fix outdated goal-agent path reference in README
• Clear build cache in quickstart scripts

---

Upgrading

git pull origin main
uv sync

# For the web frontend
cd core/frontend
npm install
npm run build

---

Contributors

Thanks to all contributors for this release!

• @Antiarin - Security hardening for agent path validation

v0.6.1

Highlights

Tool Calling Revamp

The event loop node's tool calling logic is rewritten for more reliable and predictable execution.

• New tool call routing -- restructured edge routing with improved conversation flow management
• Conversation management overhaul -- refactored conversation handling with better context tracking across iterations
• Event loop iteration fixes -- resolved iteration counting and session update spamming issues
• Pipeline visual update -- improved execution pipeline visualization in the frontend

Credential System Hardening

The credential subsystem gains a dedicated key storage layer and cleaner lifecycle management.

• Key storage module -- new key_storage.py for structured local encrypted credential management
• Migrated to ~/.hive -- credentials moved from shell config to ~/.hive directory
• Improved validation pipeline -- enhanced credential validation with better error reporting
• Session-start loading -- credentials checked and loaded into new agent sessions automatically
• Dismissable banners -- credential error banners can now be dismissed without blocking

Web Workspace Fixes

Stability and UX improvements across the frontend and server.

• SSE reconnect on session change -- stable reconnection when switching between sessions
• Tool pill per-call tracking -- individual tool call status indicators
• Cancel/pause event emission -- proper event signals for session control
• Credential modal UX -- improved add/validate/dismiss flows
• Graph visualization -- layout and node rendering refinements

---

What's New

Core Framework

• Tool call revamp -- New tool calling logic in event loop node with refactored conversation management. (core/framework/graph/event_loop_node.py, core/framework/graph/conversation.py)
• Edge routing -- New edge routing capabilities for conditional execution flow. (core/framework/graph/edge.py)
• LiteLLM improvements -- Better tool call handling and provider compatibility. (core/framework/llm/litellm.py)
• Executor updates -- Improved graph execution with better session handling. (core/framework/graph/executor.py)
• Execution stream -- Refined event streaming for tool call tracking. (core/framework/runtime/execution_stream.py)
• Key storage -- Dedicated local encrypted credential storage module. (core/framework/credentials/key_storage.py)

Frontend & Server

• SSE reconnect -- Stable reconnection on session change with tool pill per-call tracking. (core/frontend/src/hooks/use-sse.ts)
• Credential modal -- Improved add/validate/dismiss UX with dismissable error banners. (core/frontend/src/components/CredentialsModal.tsx)
• Chat panel -- Better message rendering and interaction flow. (core/frontend/src/components/ChatPanel.tsx)
• Graph visualization -- Agent graph layout and node rendering improvements. (core/frontend/src/components/AgentGraph.tsx)
• Home page -- Centered text and Open Hive branding. (core/frontend/src/pages/home.tsx)
• Server credential routes -- Updated credential and event API routes. (core/framework/server/routes_credentials.py, core/framework/server/routes_events.py)

New Tool Integrations

Tool	Description	Contributor
Intercom	Conversations, contacts, and tags management with health checker	@Ttian18
Google Analytics 4	Property listing, report generation, and GA4 data access	@Ttian18

---

Bug Fixes

• Fix event loop iteration counting
• Fix spamming session update on every iteration
• Fix dismiss credential banner not persisting
• Fix credential loading into new agent sessions
• Fix mock_mode removal from queen/coder system prompt templates
• Fix graph summary for intake nodes
• Fix incorrect CLI commands and docstrings in core docs (#5457)
• Fix wrong credential path and env var references in docs (#5458)
• Fix runtime logger test assertions for unified session run IDs (#5480)
• Fix deep research agent node initialization
• Fix SSE reconnect on session change
• Remove unused load agent code

Documentation

• Fix incorrect CLI commands and docstrings in core README
• Fix credential path and env var references in environment setup docs
• Fix configuration documentation paths
• Updated file templates and anti-patterns in hive coder reference
• Windows quickstart script update

---

Upgrading

git pull origin main
uv sync

# For the web frontend
cd core/frontend
npm install
npm run build

v0.6.0 - Open Hive 🐝

Release Notes

Release Date: February 26, 2026
Tag: v0.6.0

Open Hive

v0.6.0 is the biggest release in Hive's history. The framework breaks out of the terminal with a full web workspace -- a React SPA backed by a new FastAPI HTTP server where you can run agents, chat, visualize graphs, and manage credentials from the browser. Under the hood, a new queen/worker/judge multi-agent runtime enables agents that spawn, monitor, and coordinate other agents. The credential system gets a ground-up rewrite with local encrypted storage, health checks, and mid-session installation. And OpenAI Codex joins as a first-class LLM provider with OAuth subscription support.

---

Highlights

Web Workspace

A complete browser-based workspace replaces the TUI as the primary interface for running agents. Built with React, TypeScript, Vite, shadcn/ui, and Tailwind CSS.

• Home page -- browse agents and start sessions with an initial prompt

• Workspace -- interactive agent graph visualization (ReactFlow), live chat panel, node detail inspector, and streaming log pane

• Credentials modal -- add, validate, and revoke credentials mid-session without restarting
• Real-time streaming -- SSE-based event streaming for live execution updates
• Markdown rendering -- rich message formatting in the chat panel
• Tab persistence -- multi-session support with persistent workspace tabs

HTTP API Server

A FastAPI server at core/framework/server/ provides the backend for the web workspace and exposes a full REST + SSE API:

• Session lifecycle -- create, list, resume, cancel, reconnect sessions
• Execution control -- start, stop, send input, cancel workers
• Graph inspection -- load and query agent graph definitions
• Credential management -- validate, install, and revoke credentials at runtime
• SSE event streaming -- real-time execution events for frontend consumption
• Log queries -- structured runtime log access

Full API documentation in core/framework/server/README.md and docs/server-cli-arch.md.

Queen / Worker / Judge Multi-Agent Runtime

A new orchestration layer enables hierarchical multi-agent coordination:

• Queen lifecycle tools -- spawn, monitor, stop, and cancel worker agents with escalation ticket support
• Worker health monitoring -- stall detection, iteration tracking, and automatic cleanup of stale sessions
• Concurrent judge -- event-bus-integrated evaluation that runs alongside agent execution
• Session Manager -- replaces the old Agent Manager with full concurrent session support, reconnection, and session isolation
• 3-layer resume prompts -- robust checkpoint recovery across session restarts
• Trigger node visualization -- schedule info displayed when clicking trigger nodes in the graph

Credential System v3

The credential subsystem is rewritten again with local-first encrypted storage:

• Local credential registry (credentials/local/) -- encrypted storage with namespace support ({credential_name}/{alias})
• Health check framework -- live API validation for stored credentials (Brevo, PostgreSQL, and more)
• Mid-session credential management -- install missing credentials and resync MCP servers without restarting
• Deferred validation -- credentials validated on use rather than on load, with dismissable error banners
• Revoke and re-validate -- full lifecycle management in both TUI and web UI

OpenAI Codex LLM Provider

Codex joins as a supported LLM provider with full OAuth integration:

• OAuth consent flow -- browser pop-out for subscription authentication
• Streaming support -- through LiteLLM with proper tool call handling
• Weak model fixes -- skip auto-block when models output text instead of calling tools, remove implementation hints from judge feedback

---

What's New

Architecture & Runtime

• Web frontend -- Full React + TypeScript SPA with agent workspace, graph visualization, chat, and credential management. (core/frontend/)
• HTTP API server -- FastAPI backend with route modules for sessions, execution, graphs, credentials, events, and logs. (core/framework/server/)
• Session Manager -- Replaces Agent Manager with concurrent session support, reconnection, and isolation. (core/framework/server/session_manager.py)
• Queen lifecycle tools -- 6 tools for spawning, monitoring, stopping, and canceling worker agents with escalation support. (core/framework/tools/queen_lifecycle_tools.py)
• Worker monitoring tools -- Health checks, stall detection, iteration tracking, and stale session cleanup. (core/framework/tools/worker_monitoring_tools.py)
• Concurrent judge runtime -- Event-bus-integrated judge that evaluates agent execution in parallel. (core/framework/monitoring/judge.py)
• Codex OAuth provider -- OpenAI Codex subscription support with browser-based OAuth consent flow. (core/codex_oauth.py)
• LiteLLM overhaul -- Improved tool call handling, stop reason logging, weak model support, deprecated unused functions.
• Refactored system prompt narratives -- Cleaner prompt composition with worker path support.
• Event bus enhancements -- Structured logging, execution_id tracking, iteration-based message IDs.

TUI Improvements

• Get-started menu -- guided entry point for new users
• Ctrl+J newline fallback -- alternative newline input in chat (@Ttian18)
• Local credential add screen -- TUI modal for installing credentials locally
• Agent picker refinements -- improved loading and selection flow
• Quickstart prompt -- restart terminal prompt after quickstart setup

New Tool Integrations

Tool	Description	Contributor
Wikipedia	Search and retrieve Wikipedia article summaries	@nafiyad
arXiv	Search papers and download PDFs from arXiv	@ishaannk
PostgreSQL	Read-only SQL queries via MCP	@SANTHAN-KUMAR
Brevo	Email marketing API integration	Core team
Competitive Intelligence	Full agent template for market analysis	@nafiyad

New Agent Template

• Competitive Intelligence Agent -- Market analysis agent with web scraping, competitor tracking, and reporting. Ships as both template and ready-to-run export. (@nafiyad)

---

Bug Fixes

• Fix cancel worker logic for clean shutdown
• Fix erroneous skip judge logic that bypassed evaluation
• Fix stalled agent stop tools for proper cleanup
• Fix agent loading after graph changes
• Fix credentials modal blocking incorrectly
• Fix session reconnect and iteration-based message IDs
• Fix faster input_requested event delivery
• Fix fake goal prompt injection vulnerability
• Fix deferred credential validation with dismissable error banners
• Fix missing MCP reference in credential setup
• Fix clean up stale active sessions on worker load
• Fix credential loading and popup flow
• Fix session ID issues and duplicative starter API
• Fix isolate session loading between concurrent sessions
• Fix Codex tool call usage for streaming responses
• Fix asyncio.run crash in GraphBuilder (@vincentjiang777)
• Fix retired claude-3-5-haiku model reference with claude-haiku-4-5 (@kart1ka)
• Fix zero-config local LLM support and AgentRunner crash (@vakrahul)
• Fix Shift+Enter newline behavior in chat input (@Ttian18)
• Fix web scrape robots.txt support and networkidle wait
• Fix SuccessCriterion type field and evaluation guard (@Rudra2637)
• Reduce subprocess spawning in quickstart scripts (@mubarakar95)

Documentation

• Fix README formatting and links (@JamieJiHeonKim)
• Fix Roadmap Mermaid diagram rendering for GitHub (@sabasiddique1)
• Security tools README documentation
• Server CLI architecture documentation (docs/server-cli-arch.md)
• Worker health monitoring documentation (docs/worker-health-monitoring.md)
• HTTP API README (core/framework/server/README.md)

---

Breaking Changes

• Session Manager replaces Agent Manager -- The old agent_manager.py API surface is superseded by session_manager.py. Scripts using AgentRuntime.start_agent() should migrate to the session-based API.
• Credential namespacing -- Credentials are now stored under {credential_name}/{alias} paths. Existing credentials in ~/.hive/credentials will need to be re-stored.
• gpt-nano removed -- No longer available as an LLM option.

---

Community Contributors

A huge thank you to everyone who contributed to this release:

• Nafiyad Adane (@nafiyad) -- Wikipedia search tool and Competitive Intelligence agent template
• @ishaannk -- arXiv search and download tools
• @SANTHAN-KUMAR -- PostgreSQL read-only MCP tool
• Rudra (@Rudra2637) -- SuccessCriterion type field and evaluation guard
• Rahul Vakiti (@vakrahul) -- Zero-config local LLM support and AgentRunner crash fix
• Kartik Saini (@kart1ka) -- Retired Haiku model replacement
• @Ttian18 -- Shift+Enter newline fallback in chat input
• @mubarakar95 -- Reduced subprocess spawning in quickstart scripts (Windows)
• Jamie Kim (@JamieJiHeonKim) -- README formatting and link fixes
• Saba Siddique (@sabasiddique1) -- Roadmap Mermaid diagram fix
• Vincent Jiang (@vincentjiang777) -- asyncio.run crash fix in GraphBuilder
• Junaid (@juni2003) -- README org link fixes
• Richard Tang (@RichardTang-Aden) -- Codex subscription OAuth, credential ...

v0.5.2 - Credentials Get an Upgrade

Credentials Get an Upgrade

v0.5.2 focuses on reliability and polish. The credential system gets a full v2 rewrite with provider-based loading and runtime validation. Stripe joins the tool ecosystem as a new payment integration. Several graph-level fixes harden async safety for production workloads.

---

Highlights

Credential System v2

The credential subsystem has been rewritten with provider-based loading and a dedicated test agent. Credentials are now validated at load time with proper error typing, health checks for Google tools, and improved setup flows for both the TUI and worker agents.

# Credentials are now validated with health checks on load
hive setup-credentials

Stripe Tool Integration

New MCP tools for Stripe payment processing, enabling agents to manage payments, customers, and subscriptions programmatically.

---

What's New

Architecture & Runtime

• Credential v2 with provider loading -- Rewritten credential subsystem with provider-based loading, runtime validation, and a test agent.
• Hardened JSON parsing -- Async-safe JSON parsing for large LLM outputs, preventing corruption in concurrent node executions. (@mohammednihala123)
• Execution ID on NodeContext -- execution_id now available in NodeContext for escalate_to_coder, enabling proper execution tracking. (@Antiarin)
• Event loop node zoom in Worker Bee graph -- Worker Bee graph now renders event loop nodes with zoom support.

New Tool Integrations

Tool	Description	Contributor
Stripe	Payment processing tools for managing payments, customers, and subscriptions via Stripe API	@vakrahul

---

Bug Fixes

• Fix TUI account selection flow
• Fix ZAI API key setup
• Fix Gmail batch tool schema coercion
• Fix credential error types for clearer error reporting
• Fix Google tools health check validation
• Fix streaming output leakage
• Fix worker credential setup flow
• Harden JSON parsing for async safety and large LLM outputs (@mohammednihala123)

Documentation

• Sync zh-CN README with latest English README and fix broken links (@Schlaflied)
• Fix CLI arguments mismatch for test-debug and test-list commands (@nico-suescun)
• Add document processing recipe (@alhousseynou-ndiaye)
• Expand Available Tools table with all tools by category (@NSkogstad-AUS)
• Fix Gmail tool description inaccuracy (@NSkogstad-AUS)

---

Community Contributors

A huge thank you to everyone who contributed to this release:

• Aaryan Chandola (@Antiarin) -- Execution ID fix for escalate_to_coder
• Rahul Vakiti (@vakrahul) -- Stripe tool integration
• Nihal (@mohammednihala123) -- Hardened JSON parsing for async safety
• @Schlaflied -- Chinese README sync and link fixes
• Nicolas Suescun (@nico-suescun) -- CLI docs fix
• Alhousseynou Ndiaye (@alhousseynou-ndiaye) -- Document processing recipe
• @NSkogstad-AUS -- Tools documentation expansion

---

Upgrading

git pull origin main
uv sync

---

What's Next

• Native Web UI -- A full web interface replacing the current TUI for a richer, more accessible agent experience
• Hive Coder enhancements -- Improved code generation, smarter error recovery, and deeper framework awareness for the meta-agent
• Sub-agent structure -- First-class support for agents spawning and coordinating child agents within a graph
• GCU (Graph Compute Unit) capability -- Metered graph execution units for resource tracking and usage-based billing

v0.5.1 - The Hive Gets a Brain

Release Notes

Release Date: February 18, 2026
Tag: v0.5.1

The Hive Gets a Brain

v0.5.1 is our most ambitious release yet. Hive agents can now build other agents -- the new Hive Coder meta-agent writes, tests, and fixes agent packages from natural language. The runtime grows multi-graph support so one session can orchestrate multiple agents simultaneously. The TUI gets a complete overhaul with an in-app agent picker, live streaming, and seamless escalation to the Coder. And we're now provider-agnostic: Claude Code subscriptions, OpenAI-compatible endpoints, and any LiteLLM-supported model work out of the box.

---

Highlights

Hive Coder -- The Agent That Builds Agents

A native meta-agent that lives inside the framework at core/framework/agents/hive_coder/. Give it a natural-language specification and it produces a complete agent package -- goal definition, node prompts, edge routing, MCP tool wiring, tests, and all boilerplate files.

# Launch the Coder directly
hive code

# Or escalate from any running agent (TUI)
Ctrl+E  # or /coder in chat

The Coder ships with:

• Reference documentation -- anti-patterns, construction guide, and design patterns baked into its system prompt
• Guardian watchdog -- an event-driven monitor that catches agent failures and triggers automatic remediation
• Coder Tools MCP server -- file I/O, fuzzy-match editing, git snapshots, and sandboxed shell execution (tools/coder_tools_server.py)
• Test generation -- structural tests for forever-alive agents that don't hang on runner.run()

Multi-Graph Agent Runtime

AgentRuntime now supports loading, managing, and switching between multiple agent graphs within a single session. Six new lifecycle tools give agents (and the TUI) full control:

# Load a second agent into the runtime
await runtime.add_graph("exports/deep_research_agent")

# Tools available to agents:
# load_agent, unload_agent, start_agent, restart_agent, list_agents, get_user_presence

The Hive Coder uses multi-graph internally -- when you escalate from a worker agent, the Coder loads as a separate graph while the worker stays alive in the background.

TUI Revamp

The Terminal UI gets a ground-up rebuild with five major additions:

• Agent Picker (Ctrl+A) -- tabbed modal screen for browsing Your Agents, Framework agents, and Examples with metadata badges (node count, tool count, session count, tags)
• Runtime-optional startup -- TUI launches without a pre-loaded agent, showing the picker on first open
• Live streaming pane -- dedicated RichLog widget shows LLM tokens as they arrive, replacing the old one-token-per-line display
• PDF attachments -- /attach and /detach commands with native OS file dialog (macOS, Linux, Windows)
• Multi-graph commands -- /graphs, /graph <id>, /load <path>, /unload <id> for managing agent graphs in-session

Provider-Agnostic LLM Support

Hive is no longer Anthropic-only. v0.5.1 adds first-class support for:

• Claude Code subscriptions -- use_claude_code_subscription: true in ~/.hive/configuration.json reads OAuth tokens from ~/.claude/.credentials.json with automatic refresh
• OpenAI-compatible endpoints -- api_base config routes traffic through any compatible API (Azure OpenAI, vLLM, Ollama, etc.)
• Any LiteLLM model -- RuntimeConfig now passes api_key, api_base, and extra_kwargs through to LiteLLM

The quickstart script auto-detects Claude Code subscriptions and ZAI Code installations.

---

What's New

Architecture & Runtime

• Hive Coder meta-agent -- Natural-language agent builder with reference docs, guardian watchdog, and hive code CLI command. (@TimothyZhang7)
• Multi-graph agent sessions -- add_graph/remove_graph on AgentRuntime with 6 lifecycle tools (load_agent, unload_agent, start_agent, restart_agent, list_agents, get_user_presence). (@TimothyZhang7)
• Claude Code subscription support -- OAuth token refresh via use_claude_code_subscription config, auto-detection in quickstart, LiteLLM header patching. (@TimothyZhang7)
• OpenAI-compatible endpoint support -- api_base and extra_kwargs in RuntimeConfig for any OpenAI-compatible API. (@TimothyZhang7)
• Remove deprecated node types -- Delete FlexibleGraphExecutor, WorkerNode, HybridJudge, CodeSandbox, Plan, FunctionNode, LLMNode, RouterNode. Deprecated types (llm_tool_use, llm_generate, function, router, human_input) now raise RuntimeError with migration guidance. (@TimothyZhang7)
• Interactive credential setup -- Guided CredentialSetupSession with health checks and encrypted storage, accessible via hive setup-credentials or automatic prompting on credential errors. (@RichardTang-Aden)
• Pre-start confirmation prompt -- Interactive prompt before agent execution allowing credential updates or abort. (@RichardTang-Aden)
• Event bus multi-graph support -- graph_id on events, filter_graph on subscriptions, ESCALATION_REQUESTED event type, exclude_own_graph filter. (@TimothyZhang7)

TUI Improvements

• In-app agent picker (Ctrl+A) -- Tabbed modal for browsing agents with metadata badges (nodes, tools, sessions, tags). (@TimothyZhang7)
• Runtime-optional TUI startup -- Launches without a pre-loaded agent, shows agent picker on startup. (@TimothyZhang7)
• Hive Coder escalation (Ctrl+E) -- Escalate to Hive Coder and return; also available via /coder and /back chat commands. (@TimothyZhang7)
• PDF attachment support -- /attach and /detach commands with native OS file dialog. (@TimothyZhang7)
• Streaming output pane -- Dedicated RichLog widget for live LLM token streaming. (@TimothyZhang7)
• Multi-graph TUI commands -- /graphs, /graph <id>, /load <path>, /unload <id>. (@TimothyZhang7)
• Agent Guardian watchdog -- Event-driven monitor that catches secondary agent failures and triggers automatic remediation, with --no-guardian CLI flag. (@TimothyZhang7)

New Tool Integrations

Tool	Description	Contributor
Discord	4 MCP tools (discord_list_guilds, discord_list_channels, discord_send_message, discord_get_messages) with rate-limit retry and channel filtering	@mishrapravin114
Exa Search API	4 AI-powered search tools (exa_search, exa_find_similar, exa_get_contents, exa_answer) with neural/keyword search, domain filters, and citation-backed answers	@JeetKaria06
Razorpay	6 payment processing tools for payments, invoices, payment links, and refunds with HTTP Basic Auth	@shivamshahi07
Google Docs	Document creation, reading, and editing with OAuth credential support	@haliaeetusvocifer
Gmail enhancements	Expanded mail operations for inbox management	@bryanadenhq

Infrastructure

• Default node type → event_loop -- NodeSpec.node_type defaults to "event_loop" instead of "llm_tool_use". (@TimothyZhang7)
• Default max_node_visits → 0 (unlimited) -- Nodes default to unlimited visits, reducing friction for feedback loops and forever-alive agents. (@TimothyZhang7)
• Remove function field from NodeSpec -- Follows deprecation of FunctionNode. (@TimothyZhang7)
• LiteLLM OAuth patch -- Correct header construction for OAuth tokens (remove x-api-key when Bearer token is present). (@TimothyZhang7)
• Orchestrator config centralization -- Reads api_key, api_base, extra_kwargs from centralized ~/.hive/configuration.json. (@TimothyZhang7)
• System prompt datetime injection -- All system prompts now include current date/time for time-aware agent behavior. (@TimothyZhang7)
• Utils module exports -- Proper __init__.py exports for the utils module. (@Siddharth2624)
• Increased default max_tokens -- Opus 4.6 defaults to 32768, Sonnet 4.5 to 16384 (up from 8192). (@TimothyZhang7)

---

Bug Fixes

• Flush WIP accumulator outputs on cancel/failure so edge conditions see correct values on resume
• Stall detection state preserved across resume (no more resets on checkpoint restore)
• Skip client-facing blocking for event-triggered executions (timer/webhook)
• Executor retry override scoped to actual EventLoopNode instances only
• Add _awaiting_input flag to EventLoopNode to prevent input injection race conditions
• Fix TUI streaming display (tokens no longer appear one-per-line)
• Fix _return_from_escalation crash when ChatRepl widgets not yet mounted
• Fix tools registration problems for Google Docs credentials (@RichardTang-Aden)
• Fix email agent version conflicts (@RichardTang-Aden)
• Fix coder tool timeouts (120s for tests, 300s cap for commands)

Documentation

• Clarify installation and prevent root pip install misuse (@paarths-collab)

---

Agent Updates

• Email Inbox Management -- Consolidate gmail_inbox_guardian and inbox_management into a single unified agent with updated prompts and config. (@RichardTang-Aden, @bryanadenhq)
• Job Hunter -- Updated node prompts, config, and agent metadata; added PDF resume selection. (@bryanadenhq)
• Deep Research Agent -- Revised node implementations with updated prompts and output handling.
• Tech News Reporter -- Revised node prompts for improved output quality.
• Vulnerability Assessment -- Expanded prompts with more detailed assessment instructions. (@bryanadenhq)

---

Breaking Changes

• Deprecated node types raise RuntimeError -- llm_tool_use, llm_generate, function, router, human_input now fail instead of warning. Migrate to event_loop.
• NodeSpec.node_type defaults to "event_loop" (was "llm_tool_use")
• NodeSpec.max_node_visits defaults to 0 / unlimited (was 1)
• NodeSpec.function field removed -- FunctionNode is deleted; use event_loop nodes with tools instead.

---

Community Contributors

A huge thank you t...

v0.5.0 - 🐝 Auto Agents 🐝

Release Notes

Release Date: February 13, 2026
Tag: v0.5.0

Event-Driven Agents Are Here

v0.5.0 is our biggest architecture release yet. Hive agents can now react to the outside world -- webhooks, timers, and external events trigger agent execution alongside the interactive user session. This unlocks an entirely new class of agents: ones that don't just wait for you to talk to them, but autonomously respond to real-world events while you stay in control.

---

Highlights

Webhook & Timer Entry Points

Agents are no longer single-entry. A single agent graph can now declare multiple entry points with different trigger types -- each entering the graph at a different node.

# User-facing entry point (interactive)
EntryPointSpec(id="start", entry_node="intake", trigger_type="manual")

# External webhook triggers execution at a different node
EntryPointSpec(id="email-event", entry_node="fetch-emails",
               trigger_type="event",
               trigger_config={"event_types": ["webhook_received"]})

# Built-in scheduler -- no crontab needed
EntryPointSpec(id="email-timer", entry_node="fetch-emails",
               trigger_type="timer",
               trigger_config={"interval_minutes": 20})

Webhook server starts automatically when webhook_routes are configured. POST to http://localhost:8080/webhooks/gmail and your agent picks it up.

Timer triggers are built into the runtime -- AgentRuntime manages asyncio tasks internally. No external schedulers, no cron, no glue code.

Shared session state across entry points means the user sets up rules once via the interactive intake node, and every webhook/timer execution reuses those rules automatically.

TUI Event Source Dashboard

The TUI graph panel now displays active event sources below the agent graph:

Event Sources
  ⚡ Email Event Handler → fetch-emails
     127.0.0.1:8080/webhooks/gmail
  ⏱  Scheduled Inbox Check → fetch-emails
     every 20 min -- next in 14m 32s

Timer countdowns update live. Webhook endpoints show the exact URL to POST to. No more guessing what's listening.

Gmail Inbox Guardian -- Reference Event-Driven Agent

A fully functional Gmail automation agent ships as both a template (examples/templates/gmail_inbox_guardian/) and a ready-to-run export (exports/gmail_inbox_guardian/). It demonstrates the complete event-driven pattern:

• Intake node (client-facing, forever-alive): User defines free-text triage rules
• Fetch-emails node: Retrieves unread emails via Gmail MCP tools
• Classify-and-act node: Matches each email against rules, executes Gmail actions (star, trash, mark read, label, etc.)
• Report node: Summarizes actions taken

Three entry points run concurrently: manual rule setup, webhook-triggered triage, and a 20-minute scheduled check. Conversation memory is continuous across all of them.

---

What's New

Architecture & Runtime

• Multi-entry-point AgentRuntime -- Register webhook, timer, event, and manual entry points on a single graph. Each gets its own ExecutionStream with configurable concurrency and isolation. (#4918)
• Timer trigger type -- trigger_type="timer" with interval_minutes config. Runtime manages scheduling internally with next-fire tracking. (#4720)
• Webhook server -- Auto-starts an HTTP server for configured routes. Receives POST payloads and routes them to the correct entry point. (#4720)
• Fresh shared-session cursor clearing -- Webhook/timer executions reuse the primary session but get a fresh OutputAccumulator, preventing stale outputs from causing node skips. (#4918)
• Async LLM methods -- acomplete(), acomplete_with_tools(), _acompletion_with_rate_limit_retry() added to LiteLLMProvider. Uses litellm.acompletion and asyncio.sleep -- no more blocking the event loop. (#4920)
• Phased compaction & event bus integration -- Conversation compaction now respects phase boundaries. Event bus emits lifecycle events for all node transitions. (#4647)

Robustness & Error Handling

• Transient error retry -- Stream errors from APIConnectionError, InternalServerError, ServiceUnavailableError, and connection resets are now retried with exponential backoff instead of crashing the agent. (#4722)
• Loop detection -- Detects repeated empty responses (3+ consecutive) and breaks out instead of spinning forever.
• Recoverable stream errors -- StreamErrorEvent now carries a recoverable flag so the EventLoopNode can decide whether to retry or fail.
• Storage path validation -- Runtime validates and creates storage paths on init instead of crashing later. (#4466)

TUI Improvements

• Two-panel layout -- Graph overview on the left, chat on the right. Cleaner, more focused. (#4610)
• Event source display -- Webhooks show endpoint URLs, timers show interval and live countdown.
• Live timer countdown -- Graph panel refreshes every second to show next in Xm XXs for scheduled triggers.

New Tool Integrations

New MCP tool integrations from the community:

Tool	Description	Contributor
Google Calendar	Create, list, update calendar events	@Antiarin
Cal.com	Open-source scheduling infrastructure	@Amdev-5
BigQuery	SQL querying and data analysis	@emmanuel Nwanguma
Excel	Read/write .xlsx/.xlsm files	@amit Kumar
Google Maps Platform	6 MCP tools (geocode, directions, places, etc.)	@amit Kumar
Telegram Bot	Send messages, manage chats	@siddharth Varshney
NewsData + Finlight	Market intelligence and news aggregation	@amit Kumar
GCP Vision API	Image analysis and OCR	@T.Trinath Reddy
Gmail	List, get, trash, modify labels on emails	Core team
Time	Current date/time retrieval	@Antiarin
Security Scanner	6 passive scanning tools (SSL/TLS, HTTP headers, DNS, ports, tech stack, subdomains) + risk scorer	Core team

Example Agents

Four new ready-to-use template agents ship with this release in examples/templates/:

• Gmail Inbox Guardian -- Event-driven email triage with webhooks, timers, and user-defined rules. The user defines free-text rules ("star emails from my boss", "trash marketing newsletters"), and the agent autonomously processes incoming emails via webhook triggers or a 20-minute scheduled check. Demonstrates multi-entry-point architecture, shared session state, and continuous conversation memory. Full template + ready-to-run export. (#4925)

• Inbox Management -- Linear email management pipeline. User provides natural language rules, agent fetches inbox emails in configurable batches (up to 100), classifies each against the rules, executes Gmail actions (trash, archive, star, label, mark read/unread), and produces a summary report grouped by action type. 4-node pipeline: intake → fetch-emails → classify-and-act → report. (#4610)

• Job Hunter -- Resume-driven job search and outreach agent. Analyzes your resume to identify strongest role fits, searches the web for 10 matching job listings, lets you interactively select which to pursue, then generates tailored resume customization notes and cold outreach emails for each selected position -- all saved to disk. 4-node pipeline with 3 client-facing interactive nodes: intake → job-search → job-review → customize. (#4759)

• Passive Vulnerability Assessment -- OSINT-based website security scanner. Accepts a target domain and runs 6 passive scanning tools (SSL/TLS, HTTP headers, DNS security, port scanning, tech stack detection, subdomain enumeration), produces letter-grade risk scores (A-F) per category with a weighted overall grade, presents findings at an interactive checkpoint where the user can request deeper scanning or generate the report, and delivers a self-contained HTML risk dashboard with color-coded grades, detailed findings, and developer-focused remediation steps. 5-node pipeline with feedback loop: intake → passive-recon → risk-scoring → findings-review ↔ passive-recon → final-report ↻ intake. (#4922)

---

Bug Fixes

• Fix call_from_thread errors in TUI when events fire from executor threads
• Fix stale memory filtering -- webhook executions no longer inherit outdated nullable outputs
• Fix conversation history persistence causing fetch-emails to be skipped on subsequent triggers
• Fix default batch size (50 → 10) to prevent context overflow on large inboxes
• Fix unicode bullet rendering on Windows (#4525)
• Fix uv local storage path resolution
• Remove unused send_budget_alert_email tool, require explicit provider param

Documentation

• Fix incorrect directory names in recipes README (@zhanglinqian)
• Fix typos in roadmap.md (@LukeM94)
• Fix nonexistent setup-python.sh references (@Hima-de, @jaathavan18)
• Fix broken exports/ link in environment-setup.md (@jaathavan18)

---

Breaking Changes

None. All changes are backwards-compatible. Agents without event sources continue to work exactly as before.

---

Windows

• Windows Defender exclusions for 40% faster uv sync (@e-cesar9)
• Improved failure messaging with success rate display (@e-cesar9)
• Path validation for Defender exclusions (@e-cesar9)
• Unicode bullet replacement with ASCII dashes (@richard Tang)

---

Community Contributors

A huge thank you to everyone who contributed to this release:

• Amit Kumar (@Amdev-5) -- Cal.com, Excel, Google Maps, NewsData + Finlight integrations
• Pravin Mishra ...

v0.4.4 🐝

Release Date: February 12, 2026
Tag: v0.4.4

🧵 Continuous Agents & Multi-IDE Support

This release introduces the "Continuous Agent" pattern for preserving conversation context across nodes, adds first-class support for Codex CLI, Antigravity IDE, and Opencode, and brings Windows support with PowerShell quickstart.

✨ What's New

🧠 Continuous Agent Memory System

One conversation, many phases. Agents can now thread a single continuous conversation across their entire graph — preserving reasoning chains, tool history, and user intent across phase boundaries.

• Conversation Threading — When conversation_mode="continuous", one NodeConversation flows across all event_loop nodes
• Cumulative Tools — Tools accumulate across node transitions (new phases ADD tools, never remove)
• Layered Prompt Architecture — Three-layer "onion" system: Identity → Narrative → Focus
• Phase-Graduated Compaction — Smart memory management that respects phase boundaries
• Level 2 Conversation-Aware Judge — Evaluates whether conversation meets success criteria, not just checkboxes

🤖 Codex CLI Integration

First-class Codex support. Build Hive agents using OpenAI's Codex CLI with shared MCP servers and skills.

• Project-level Codex configuration in .codex/config.toml
• Shared skill mountpoints via symlinks to existing Hive skills
• Quickstart bootstrap and verification for Codex setup
• GitHub workflow for Codex-powered issue triage

🌌 Antigravity IDE Support

Google's AI-powered IDE joins the party. Full MCP server and skill integration following the same pattern as Cursor.

• .antigravity/mcp_config.json for agent-builder and tools servers
• Symlinked skills from .claude/skills/ for single source of truth
• Comprehensive setup guide with troubleshooting

🖥️ Opencode Integration

Native Opencode support with cross-platform compatibility. Full support for Windows (handles ; paths), macOS, and Linux.

• MCP server integration for agent-builder and tools
• /hive command access to all Hive skills
• Verified on Windows 11

🪟 Windows Support

First-class Windows experience. PowerShell quickstart script and CLI compatibility.

• quickstart.ps1 — Full Windows setup script with PS 5.1 compatibility
• Fixed path handling for Windows absolute paths
• Environment loader compatibility tweaks

🔬 SerpAPI Tools for Research

Academic and patent research capabilities. Five new tools powered by SerpAPI for Google Scholar and Google Patents.

• scholar_search — Search papers with citation counts, author IDs, PDF links
• scholar_get_citations — Get formatted citations (MLA, APA, Chicago, Harvard, Vancouver) + BibTeX
• scholar_get_author — Author profiles with h-index, i10-index, research interests
• patents_search — Search patents with country/status/date filters
• patents_get_details — Detailed patent info by publication number

🧪 Automated Testing Skill

Test your agents like production code. New hive-test skill with proper runtime fixtures.

• Unified runtime path — All agents flow through AgentRuntime
• runner fixture — Session-scoped async runner matching real hive run path
• auto_responder fixture — Auto-inject responses to client-facing nodes
• Headless session resume with --resume-session and --checkpoint flags

🐛 Bug Fixes

• TUI Multiline Input — Replaced single-line Input with TextArea; Shift+Enter for newlines, multiline paste support (#4423, #4451)
• Cross-Platform Clipboard — Added Windows (clip.exe) and Linux (xsel) clipboard fallbacks (#4423)
• Windows Path Security — Fixed unix-style absolute path handling on Windows (#1204)
• MCP Error Logging — Errors in _load_active_session now logged instead of silenced (#682)
• LLM Token Counting — Fixed token counting to include validation retries (#4299)
• Codex Prompt Injection — Fixed prompt injection risk in issue triage workflow
• PowerShell Compatibility — Fixed Join-Path and parser errors for PS 5.1 (#4437)
• Gemini API Key — Fixed env var from GOOGLE_API_KEY to GEMINI_API_KEY
• Stale Nullable Outputs — Nullable output keys now cleared on node re-visit
• All-Nullable Implicit Judge — Returns RETRY instead of ACCEPT when all nullable keys unset

📚 Documentation

• New Autonomous Agent Guide for building self-directed agents
• Developer Success documentation and roadmap
• Fixed environment setup docs for uv workspace
• Fixed .env.example references in tools README

🙏 Contributors

Special thanks to our community contributors:

• @Antiarin (Aaryann Chandola) — Codex CLI integration (#4587)
• @himanshu748 (Himanshu Kumar) — Antigravity IDE support (#2862)
• @Ttian18 (Zhang) — TUI multiline input & clipboard (#4451)
• @Amdev-5 (Amit Kumar) — SerpAPI tools for Scholar & Patents (#3986)
• @vakrahul — Opencode integration (#4243)
• @YashovardhanB28 (Yashovardhan Bangur) — Token counting fix (#4299)
• @GovindhKishore — PowerShell syntax fix (#4437)
• @jaathavan18 — Documentation cleanup (#4538, #4520, #4498)
• @nhockcuncon77 (Minh Viet Huynh) — Contributing docs fix (#4287)
• @e-cesar9 — Typo fixes and logging improvements (#4479, #4417)
• @saurabh007007 — Spelling fixes (#4441)
• @NamanRajput-git — Submission email fix (#4021)
• @AkhileshBabuT — Environment setup fix (#3894)

---

What's Next?

We're working on:

• Interactive debugging UI for checkpoint inspection
• Hot reloading support for code changes
• Graph version compatibility checking
• Enhanced agent observability dashboard

Stay tuned for v0.4.5! 🚀