Skip to content

agent-tools.md

Latest commit

 

History

History
58 lines (40 loc) · 3.36 KB

agent-tools.md

File metadata and controls

58 lines (40 loc) · 3.36 KB

Agent Tools

Myco provides two categories of tools: MCP tools that agents call programmatically, and skills that users invoke via slash commands. Both are available automatically when Myco is installed.

Automatic Context Injection

Before any tool is called, Myco injects context at two points automatically:

  • Session start (SessionStart hook) — the digest extract is injected, giving the agent a pre-computed understanding of the project before it asks a single question.
  • Per-prompt (UserPromptSubmit hook) — relevant spores are retrieved via vector search and injected, providing targeted context for the current task.

See the Lifecycle docs for the full event flow.

MCP Tools

12 tools exposed via the Model Context Protocol to any compatible agent runtime. Agents discover them automatically through MCP.

Search & Recall

Tool Purpose
myco_search Semantic + keyword search across sessions, spores, and plans. Accepts a query string and optional type filter (spore, session, plan).
myco_recall Context retrieval based on git branch and files. Returns relevant spores, session history, and plan progress for the current work.
myco_context On-demand digest extract at a specific token tier (1500, 3000, 5000, or 10000).

Knowledge Capture

Tool Purpose
myco_remember Save an observation as a spore. Types: gotcha, decision, discovery, trade_off, bug_fix.
myco_supersede Mark an older spore as replaced by a newer one. The old spore is preserved with lineage metadata but filtered from search and digest.
myco_consolidate Merge related spores into a single wisdom note. Source spores are marked superseded with bidirectional links.

Browsing & Navigation

Tool Purpose
myco_sessions Browse session history with filters: branch, plan, user, or date range.
myco_plans List active plans and their progress, or read a specific plan by ID.
myco_graph Traverse vault connections via wikilinks in either direction, with configurable depth.
myco_orphans Find vault notes with no incoming or outgoing wikilinks.
myco_team See teammate activity, filtered by files or plan.

Debugging

Tool Purpose
myco_logs View daemon logs filtered by level and component (daemon, processor, digest, intelligence, consolidation, etc.).

Skills

Skills are user-invocable slash commands that provide guided workflows. Type the command in your agent's prompt to activate.

Command Purpose
/myco-setup Initialize Myco in a new project. Guides first-time setup for the vault, LLM provider, and embedding model. Run once per project.
/myco The primary skill for ongoing work. Use when making design decisions, debugging non-obvious issues, encountering gotchas, or needing context about prior work. Provides guidance on when and how to use each MCP tool, and patterns for vault hygiene (superseding stale spores, consolidating related observations).
/myco-rules Create, audit, or improve project rules files (CLAUDE.md, AGENTS.md). Helps write specific, enforceable rules that agents actually follow. Also triggered when Myco detects recurring patterns that should become project rules.