Skip to content

features.md

Latest commit

 

History

History
322 lines (248 loc) · 16.2 KB

features.md

File metadata and controls

322 lines (248 loc) · 16.2 KB
Raw

MCP Explorer — Feature Reference

Features grouped by navigation section and menu item.

🌐 Global / App Shell

  • Collapsible sidebar — toggle the navigation panel to reclaim screen space; nav items show icon-only tooltips when collapsed
  • Navigation groups — menu items are grouped into colour-coded sections (Infrastructure, Configuration, MCP Explorer, Testing)
  • Command palette (⌘K / Ctrl+K) — fuzzy-search and jump to any page instantly from the keyboard
  • Theme switcher — choose from 10 built-in themes; selection is persisted per user
    • Command Dark, Command Light
    • Nord, Dracula, Catppuccin Mocha
    • Solarized Light
    • GitHub Dark, GitHub Light
    • Material Dark, Material Light
  • Version pills — topbar shows live .NET runtime version, frontend version, and API version
  • Active connections badge — topbar indicator showing how many MCP connections are currently live
  • Active dev tunnels badge — topbar indicator showing how many Dev Tunnels are currently online

🚇 Dev Tunnels

Tunnel Management

  • Create / start / stop / delete Dev Tunnels from the UI with named configurations and access mode selection
  • Anonymous caller access — anonymous tunnels use the devtunnel host -p <port> -a flow so external services (Stripe, GitHub, your MCP tool) can POST without a token; the CLI host itself still signs in once
  • Device-code login dialog — trigger devtunnel CLI authentication from the UI and stream login instructions live
  • CLI availability state — if the CLI is missing or misconfigured, the UI shows a clear unavailable state instead of failing with a generic server error
  • Multiple tunnel support — manage several named tunnels at once
  • Copy webhook URL — one-click copy of the public callback endpoint

Live Inspection

  • SSE-powered event tape — incoming webhook events animate into the inspector in real time
  • Per-tunnel stream state — connected / connecting / disconnected indicator for live capture
  • Event detail panels — inspect parsed JSON body, raw payload, and captured headers
  • Sensitive header masking — Authorization, cookies, and API-key-style headers are redacted before display

Time Travel

  • Persisted event history — captured events are stored per tunnel and survive restarts
  • Timeline scrubber — jump back through event history with a histogram-backed playback rail
  • Replay captured webhook — resend a stored request to any target URL and inspect the downstream response

Tool Integration

  • Webhook URL smart-fill — tool parameters named like webhookUrl / callbackUrl can be populated from an active tunnel directly in the Tools view

🔌 Infrastructure › Connections

Connection Management

  • Create a new connection with name, endpoint URL, optional description, and group assignment
  • Edit an existing connection inline (all fields including name)
  • Delete a connection with confirmation dialog
  • Rename a connection (name change updates all references)
  • Group assignment — assign each connection to a named group for organisation
  • Connection list — sorted and filterable; supports search by name/endpoint

Connection Groups

  • Create / edit / delete groups — each group has a name, colour, and optional description
  • Rename a group — cascades the rename to all member connections automatically
  • Group filter — click a group chip to filter the connections list

Connection Display Options

  • Show/hide timestamps — toggle created/updated timestamps in the connection list
  • Sort order — change the order connections are listed
  • Group view toggle — switch between grouped and flat connection views

Health & Connectivity

  • Connection health badge — live status indicator (healthy / degraded / unknown)
  • Manual reconnect — button to reconnect a degraded connection without leaving the page

Favourites

  • Star individual connections as favourites
  • Show favourites first toggle — persisted preference

Export

  • Select connections to export — multi-select dialog with text filter to find specific connections
  • Select / deselect all checkbox for bulk selection
  • Password protection — enter a password before exporting; AES-256-GCM encryption with PBKDF2-SHA256 key derivation
  • Auto-generate password — cryptographically random 20-character password (upper, lower, digits, symbols)
  • Copy password to clipboard — with 2-second "Copied!" feedback
  • Download as connections-export.json

Import

  • Upload exported file — drag-and-drop or file picker
  • Password prompt — enter the export password; clear error if the password doesn't match
  • Collision handling — if an imported connection name already exists, it is automatically renamed (v2), (v3), etc.

⚙️ Configuration › AI Models

  • Create / edit / delete LLM model definitions with confirmation on delete
  • Rename model (name change with collision detection — returns 409 if new name already exists)
  • Provider types supported: OpenAI, Azure OpenAI, Azure AI Foundry, Ollama, Anthropic, Custom
  • Deployment name field shown for Azure OpenAI and Azure AI Foundry providers
  • Endpoint URL field for providers that support a custom base URL
  • API key field with show/hide toggle (PrimeVue Password component)
  • Masked API key display in the model list (shows only last 4 characters)
  • Set as default — mark one model as the default for Chat and Prompts
  • Default indicator badge in the model list

🛡️ Configuration › Data Guard

  • Additional sensitive field keys — define extra JSON property names (beyond the built-ins) that should be masked in the Chat view
  • Custom regex patterns — add free-text regular expressions to redact arbitrary patterns from chat messages
  • Allowed fields — whitelist specific field names to exempt them from masking even if they match a sensitive pattern
  • Add / remove individual entries from each list
  • Save configuration to the backend (persisted in settings.json)
  • Live effect — chat masking picks up the saved config on next page load

🔧 MCP Explorer › Tools

Connection & Tool Selection

  • Connection selector — pick an active MCP connection from a dropdown
  • Tool list — alphabetically sorted list of all tools exposed by the selected connection
  • Search / filter — live text filter to narrow the tool list
  • Tool badge count — shows how many tools match the current filter vs total
  • Warning badge — amber badge when the filter is active and results are reduced

Favourites

  • Star individual tools — toggle favourite per tool; persisted to the backend
  • Show favourites first toggle — floats starred tools to the top; persisted

Tool Invocation

  • Dynamic parameter form — auto-generated input fields from the tool's JSON Schema
  • Required field validation — lists missing required fields before invoking
  • Invoke button with loading state and label feedback ("Invoking…")
  • Retry on exhaustion — "Retry" banner appears if all retry attempts are spent
  • Server-side elicitation — if the MCP server sends an elicitation request mid-invocation, a modal dialog appears for the user to respond

Parameter History

  • Per-tool history — up to 5 recent parameter sets are stored per tool
  • Load a previous run — click a history entry to populate the form with past values
  • Delete a run — remove a history entry with a confirmation dialog (hover to reveal trash icon)
  • Per-field history popover — click the history icon next to any field to see and re-select previously entered values for that field specifically

Documentation

  • Single tool docs — markdown-rendered documentation for the selected tool, including input schema and output schema
  • Bulk tool docs — view documentation for all visible (filtered) tools in one dialog

Connection Health

  • Health indicator per connection chip
  • Reconnect button per connection chip for quick re-establishment

📝 MCP Explorer › Prompts

Prompt Discovery

  • Connection selector — pick from active MCP connections
  • Prompt list — alphabetically sorted; live text search/filter

Favourites

  • Star individual prompts as favourites; persisted
  • Show favourites first toggle; persisted

Prompt Execution

  • Dynamic argument form — input fields generated from the prompt's argument schema
  • Execute — submit arguments and display the rendered prompt messages
  • Inline LLM response — send the executed prompt directly to a configured LLM model without leaving the page
  • Model picker — popover to select the target LLM model (loads configured models)
  • Streaming response — LLM reply streams token-by-token into the inline response panel
  • Cancel streaming — abort button while the LLM is responding
  • Result tabs — view the Prompt response and the LLM response in separate tabs

Documentation

  • Single prompt docs — markdown documentation for the selected prompt
  • Bulk prompt docs — documentation for all visible prompts in one dialog

🗄️ MCP Explorer › Resources

Resource Discovery

  • Connection selector — pick from active MCP connections
  • Resource list — alphabetically sorted; live text search/filter

Favourites

  • Star individual resources as favourites; persisted
  • Show favourites first toggle; persisted

Resource Reading

  • Read resource — fetch and display the resource content
  • Content rendering — text/plain, text/markdown (rendered), application/json (pretty-printed)

Documentation

  • Single resource docs — markdown documentation for the selected resource
  • Bulk resource docs — documentation for all visible resources

📋 MCP Explorer › Templates (Resource Templates)

Template Discovery

  • Connection selector — pick from active MCP connections
  • Template list — alphabetically sorted; live text search/filter

Favourites

  • Star individual templates as favourites; persisted
  • Show favourites first toggle; persisted

Template Reading

  • Dynamic URI parameter form — input fields derived from the template's URI pattern
  • Read template — substitute values into the URI template and fetch the resource
  • Content rendering — same as Resources (plain, markdown, JSON)

Documentation

  • Single template docs — markdown documentation for the selected template
  • Bulk template docs — documentation for all visible templates

🔔 MCP Explorer › Elicitations

  • Pending elicitation list — shows all server-initiated elicitation requests awaiting a response
  • Elicitation details dialog — view the full request content with its schema
  • Respond / Decline — submit a response or decline the elicitation from the history list
  • Live elicitation dialog — when a tool invocation triggers an elicitation, a modal appears inline on the Tools or Chat page for immediate interaction
  • Dynamic form rendering:
    • String, number, boolean fields
    • Enum with ≤ 3 options → radio/checkbox card layout
    • Enum with > 3 options → Select (single) or MultiSelect (multi) dropdown
  • Elicitation history — past elicitations are shown with their status (pending / responded / declined)

🔀 Testing › Workflows

Workflow Management

  • Create / edit / delete workflow definitions; delete requires confirmation
  • Multi-step pipeline — add, remove, and reorder steps with up/down arrows
  • Step renumbering — automatic when steps are added, removed, or moved

Workflow Steps

  • Tool selector — per-step dropdown of tools from the workflow's default connection
  • Parameter mapping — map input parameters from:
    • A previous step's output (with JSON path selector)
    • A manually entered value
    • A runtime prompt (user is asked at execution time)
  • JSON path browser — dialog to browse the output schema of a previous step and select a path
  • Array iteration mode — per-step setting: None, Each (iterate), First, Last
  • Error handling mode — Stop on Error or Continue on Error

Execution

  • Execute workflow — run the full pipeline against a selected connection
  • Runtime parameter prompts — if any step has prompt-at-runtime mappings, a dialog collects them before execution starts
  • Execution results — per-step output displayed in result tabs
  • Execution history — recent runs stored and accessible per workflow

Load Testing

  • Configure load test — set duration (seconds) and parallelism (concurrent workers)
  • Run load test — execute the workflow repeatedly under load
  • Live progress dialog — shows elapsed time, request count, and in-flight workers during the test
  • Results chart — bar/line chart of throughput and latency per time bucket
  • Load test history — previous load test runs stored per workflow

Import / Export

  • Export workflow — download a workflow definition as a JSON file
  • Import workflow — upload a JSON file; name collisions are resolved automatically by appending (v2), (v3), etc.

💬 Testing › Chat

Sessions

  • Multiple chat sessions — create and switch between named conversation sessions in the sidebar
  • Rename session — double-click or use the rename dialog
  • Delete session — remove a session with confirmation
  • Session sidebar — keyboard-navigable list of sessions (↑/↓ arrows)

Messaging

  • Send messages — Enter to send, Shift+Enter for a new line
  • Select LLM model — pick from configured AI models per session
  • System prompt — set and edit a custom system prompt per session
  • Streaming responses — assistant replies stream token-by-token

Tool Calling

  • Automatic tool call display — tool calls made by the LLM are shown as collapsible blocks
  • Tool call parameters — expand to see the JSON parameters sent; pretty-printed
  • Tool call results — expand to see the raw result returned
  • Sensitive parameter masking — tool call parameters matching Data Guard rules are automatically redacted with ████
  • Reveal masked values — 👁 toggle to show the original values

Prompt Invocation

  • Prompt picker — search and select a prompt from any active connection via a dedicated dialog
  • Prompt parameter form — fill in prompt arguments before injecting
  • Prompt invocation block — chat messages show the prompt name, connection, and parameter summary; expandable

Slash Commands

  • / trigger — type / in the message box to open the slash command menu
  • Fuzzy search — filter commands by typing after /
  • Keyboard navigation — ↑/↓ to move, Enter or Tab to select, Escape to dismiss
  • Built-in commands: /clear (clear chat history), /model (change model), /system (set system prompt), /stats (show token usage stats)

Sensitive Data Masking (User Messages)

  • Auto-detect sensitive content — user messages are scanned against Data Guard patterns before display
  • 🔒 Contains sensitive values banner — shown on user message bubbles that were redacted
  • 👁 Reveal toggle — show the original unredacted text per message

Statistics & Info

  • Token usage stats dialog — shows prompt tokens, completion tokens, total tokens, and thinking time across the session
  • Per-message metadata — timestamp, model name, and response time shown for each assistant message
  • Thinking time display — shows how long the model spent "thinking" (for models that support it)

Report

  • Export chat report — generate a formatted report of the conversation (ChatReportDialog)

Markdown Rendering

  • Markdown-rendered responses — assistant messages are rendered with full markdown (headings, code blocks, lists, tables, bold/italic)
  • Toggle raw/rendered — switch between rendered markdown and raw text per session
  • Copy message — copy the raw text of any message to the clipboard

This document is auto-maintained. Update when adding new features.