API Reference

Implementation References

API entrypoints: fetch-app/src/index.ts, fetch-app/src/api/status.ts.
Tool endpoints/handlers: fetch-app/src/tools/*, fetch-app/src/validation/tools.ts.
Supporting runtime services: fetch-app/src/security/*, fetch-app/src/utils/logger.ts, fetch-app/src/utils/version.ts.
Validation tests: fetch-app/tests/unit/status-api.test.ts, fetch-app/tests/unit/*-tools.test.ts.

flowchart LR
    Client["Manager / Browser / cURL"] --> API["Bridge API :8765"]
    API --> Status["status + health endpoints"]
    API --> Admin["admin endpoints (logout, reload, sessions)"]
    Admin --> Auth{"Bearer ADMIN_TOKEN"}
    Auth -->|valid| Actions["Execute action"]
    Auth -->|invalid| Reject["401/403"]

Status API

The Bridge exposes an HTTP API on port 8765.

Endpoint Summary

Method	Path	Auth	Purpose
`GET`	`/api/status`	none	Current bridge status payload
`GET`	`/api/health`	none	Lightweight health probe
`POST`	`/api/logout`	bearer token	Disconnect WhatsApp session
`POST`	`/api/whatsapp/start`	bearer token	Start WhatsApp bridge session (on-demand setup)
`POST`	`/api/whatsapp/restart`	bearer token	Restart WhatsApp bridge session (fresh QR cycle)
`POST`	`/api/config/reload`	bearer token	Reload mounted `.env` values into runtime env
`GET`	`/api/sessions`	bearer token	List sessions (summary view)
`GET`	`/api/sessions/:id`	bearer token	Retrieve one full session
`DELETE`	`/api/sessions/:id`	bearer token	Delete one session
`POST`	`/api/sessions/:id/clear`	bearer token	Clear one session history

Session ID grammar for :id: ^[A-Za-z0-9_-]+$

GET /api/status

Returns system health and WhatsApp connection state.

Response:

{
  "state": "authenticated",
  "qrCode": null,
  "qrUrl": null,
  "uptime": 3600,
  "messageCount": 42,
  "lastError": null,
  "version": "0.0.94",
  "notificationMetrics": {
    "total": 120,
    "templateEphemeral": 70,
    "llmRewriteSuccess": 28,
    "templateFallback": 22,
    "rewriteAttempts": 50,
    "rewriteDisabled": 0,
    "rewriteTimeouts": 1,
    "rewriteErrors": 3,
    "sanitizeRejects": 4,
    "duplicateSuppressions": 12
  },
  "responseFormattingMetrics": {
    "normalizedCount": 18,
    "chunkedCount": 3,
    "fallbackSplitCount": 1
  }
}

Field	Type	Description
`state`	string	`initializing`, `qr_pending`, `authenticated`, `disconnected`, `error`
`qrCode`	string\|null	Raw QR code data when `state` is `qr_pending`
`qrUrl`	string\|null	URL to render QR code image when `state` is `qr_pending`
`uptime`	number	Seconds since start
`messageCount`	number	Messages processed this session
`lastError`	`string`\|`null`	Most recent error message
`version`	`string`	Running application version
`notificationMetrics`	`object`	Notification rewrite/template counters for operational telemetry
`responseFormattingMetrics`	`object`	WhatsApp response formatting/chunking counters (`normalizedCount`, `chunkedCount`, `fallbackSplitCount`)

GET /api/sessions

Returns summarized session metadata.

Headers:

Authorization: Bearer <ADMIN_TOKEN>

Response:

{
  "success": true,
  "sessions": [
    {
      "id": "ses_abc123",
      "userId": "15551234567@c.us",
      "messageCount": 42,
      "lastActivityAt": "2026-02-13T15:40:00.000Z",
      "createdAt": "2026-02-13T14:00:00.000Z",
      "activeProject": "fetch-app"
    }
  ]
}

GET /api/sessions/:id

Returns the full message history for a specific session.

Headers:

Authorization: Bearer <ADMIN_TOKEN>

Response:

{
  "success": true,
  "session": {
    "id": "ses_1739572800",
    "messages": [
      {
        "role": "user",
        "content": "List my workspaces",
        "timestamp": "2024-02-14T22:40:00Z"
      },
      {
        "role": "assistant",
        "content": "You have 3 workspaces: fetch-bridge, fetch-manager, and kennels.",
        "timestamp": "2024-02-14T22:40:02Z"
      }
    ]
  }
}

Field	Type	Description
`id`	string	Session ID
`messages`	array	List of message objects
`messages[].role`	`string`	`user`, `assistant`, or `tool`
`messages[].content`	`string`	Raw message content
`messages[].timestamp`	`string`	ISO 8601 timestamp

GET /api/health

Lightweight health check (used by the Go TUI manager).

Response: { "healthy": true }

POST /api/logout

Disconnects the WhatsApp session. Requires authentication.

Headers:

Authorization: Bearer <ADMIN_TOKEN>

Response: { "success": true }

POST /api/config/reload

Reloads mounted .env values into process environment. Requires authentication.

Headers:

Authorization: Bearer <ADMIN_TOKEN>

Response:

{
  "success": true,
  "updatedKeys": ["FETCH_RATE_LIMIT_MAX", "ENABLE_BROWSER"],
  "message": "2 key(s) updated"
}

DELETE /api/sessions/:id

Deletes a session by ID. Requires authentication.

POST /api/sessions/:id/clear

Clears message history for one session. Requires authentication.

The ADMIN_TOKEN can be auto-generated on startup when not set explicitly. For stable manager/TUI session operations, set ADMIN_TOKEN in .env so the manager and bridge always use the same token across restarts.

Orchestrator Tools

These are the registered tools available to the LLM during the ReAct loop. They are defined with Zod schemas in src/validation/tools.ts and registered in src/tools/registry.ts.

Narrative Outputs: All tool handlers return human-readable narrative text in their output field (consumed by the LLM) with full structured data in the metadata field (used for session state sync). This improves LLM reasoning compared to raw JSON dumps.

Tool Module Ownership

Use this table when updating behavior so docs stay aligned with the implementation.

Tool Category	Source Module
Workspace (`workspace_*`, `file_delete`, `folder_delete`)	`fetch-app/src/tools/workspace.ts`
Task (`task_*`)	`fetch-app/src/tools/task.ts`
Interaction (`ask_user`, `report_progress`)	`fetch-app/src/tools/interaction.ts`
GitHub (`github_*`)	`fetch-app/src/tools/github.ts`
Web (`web_fetch`, `web_search`)	`fetch-app/src/tools/web.ts`
Browser (`browser_*`)	`fetch-app/src/tools/browser.ts`
Workflow/Cron/Runtime (`workflow_`, `cron_`, `app_run`, `app_test`, `browser_test`)	`fetch-app/src/tools/workflow.ts`

The registry entry point is fetch-app/src/tools/registry.ts, and input schemas are defined in fetch-app/src/validation/tools.ts.

Supporting Module Ownership

Responsibility	Source Module
Shared validation primitives (ids, paths, limits)	`fetch-app/src/validation/common.ts`
Tool input schemas (canonical tool-name/arg contract)	`fetch-app/src/validation/tools.ts`
Kennel Docker command execution	`fetch-app/src/utils/docker.ts`
Runtime version lookup	`fetch-app/src/utils/version.ts`
Task/progress id generation	`fetch-app/src/utils/id.ts`
Bridge logging + log-level filtering	`fetch-app/src/utils/logger.ts`
Envelope response contract + renderer	`fetch-app/src/agent/envelope.ts`, `fetch-app/src/agent/composer.ts`
Proactive task message composition	`fetch-app/src/bridge/progress-message.ts`
Voice-note transcription (`whisper-cpp`)	`fetch-app/src/transcription/index.ts`
Image analysis (vision model calls)	`fetch-app/src/vision/index.ts`
Workspace lifecycle + git/GitHub operations	`fetch-app/src/workspace/manager.ts`
Project profile detection (framework/package manager/test runner)	`fetch-app/src/workspace/profiler.ts`
Repository map generation (context summary)	`fetch-app/src/workspace/repo-map.ts`
Language symbol extraction for repo-map	`fetch-app/src/workspace/symbols.ts`
Workspace domain contracts	`fetch-app/src/workspace/types.ts`
Bridge bootstrap/shutdown orchestration	`fetch-app/src/index.ts`

Workspace Tools (9)

workspace_list

List all projects in the workspace directory.

Parameters: none (empty object)

Returns: Narrative text (e.g. "3 workspaces: my-app (active, TypeScript, main), api (Go, dev)") with full structured data in metadata.

Danger Level: SAFE

workspace_select

Switch the active project. Triggers a system prompt rebuild so the LLM sees the new project context.

Parameters:

Name	Type	Required	Description
`name`	string	✅	Workspace name to select

Returns: Narrative text (e.g. "Switched to my-app (TypeScript, on main)") with full workspace data in metadata.

Danger Level: SAFE

workspace_status

Get detailed workspace status including git info.

Parameters:

Name	Type	Required	Description
`name`	string	—	Workspace name (uses active workspace if not specified)

Returns: Narrative text (e.g. "my-app (TypeScript) - on feature-x - 3 modified, 1 untracked - 2 ahead") with full status in metadata.

Danger Level: SAFE

workspace_create

Create a new workspace/project. Automatically creates a GitHub repository and pushes initial commit if GH_TOKEN is configured.

Parameters:

Name	Type	Required	Description
`name`	string	✅	Project name (alphanumeric, hyphens, underscores, max 64 chars)
`template`	string	—	Template: `empty`, `node`, `python`, `rust`, `go`, `react`, `next` (default: `empty`)
`description`	string	—	Brief description of the project (max 256 chars)
`initGit`	boolean	—	Initialize a git repository (default: `true`)

Returns: Narrative text (e.g. "Created my-app (TypeScript, npm, git initialized)") with full workspace data in metadata.

Danger Level: MODERATE

workspace_delete

Delete a workspace permanently. Requires explicit confirmation.

Parameters:

Name	Type	Required	Description
`name`	string	✅	Workspace to delete
`confirm`	boolean	✅	Must be `true` to confirm deletion

Returns: Narrative text (e.g. "Deleted workspace old-project")

Cannot delete the currently active workspace. Select a different workspace first.

Danger Level: DANGEROUS

file_delete

Delete a specific file within the workspace. Use this for simple file removal or deleting untracked files. Requires explicit user confirmation.

Parameters:

Name	Type	Required	Description
`path`	string	✅	Relative path to the file to delete
`workspace`	string	—	Workspace name (uses active workspace if not specified)
`confirm`	boolean	✅	Must be `true` to confirm deletion

Returns: Narrative text (e.g. "Deleted file src/temp.txt")

Danger Level: DANGEROUS

folder_delete

Delete a directory and all its contents. Use this for recursive folder removal. Requires explicit user confirmation.

Parameters:

Name	Type	Required	Description
`path`	string	✅	Relative path to the folder to delete
`workspace`	string	—	Workspace name (uses active workspace if not specified)
`confirm`	boolean	✅	Must be `true` to confirm deletion

Returns: Narrative text (e.g. "Deleted folder src/generated").

Danger Level: DANGEROUS

workspace_sync

Sync workspace to GitHub. Stages changes, commits, creates repo if needed, and pushes.

Parameters:

Name	Type	Required	Description
`name`	string	—	Workspace to sync (uses active workspace if not specified)
`message`	string	—	Commit message (auto-generated from changes if not provided, max 256 chars)

Returns: Narrative text (e.g. "Committed 4 files: 'Add auth' (abc1234), pushed to origin/main") with sync details in metadata.

Danger Level: MODERATE

workspace_publish

Create a new GitHub repository from an existing workspace and push all commits.

Parameters:

Name	Type	Required	Description
`name`	string	—	Workspace to publish (uses active workspace if not specified)
`description`	string	—	Description for the new GitHub repository (max 256 chars)
`isPublic`	boolean	—	Make the repository public (default: `false` / private)

Returns: Narrative text (e.g. "Published my-app to github.com/user/my-app (private)") with repo URL in metadata.

Fails if the workspace already has a remote. Use workspace_sync to push changes to an existing repo.

Danger Level: MODERATE

Task Tools (4)

task_create

Create and start a new coding task for complex work (refactoring, features). NEVER use this for simple file deletion or single-file removal. Delegates to a harness (Claude/Gemini/Copilot/OpenCode/Codex) running in the Kennel container via docker exec.

Parameters:

Name	Type	Required	Description
`goal`	string	✅	Clear description of what to accomplish
`agent`	string	—	Agent to use: `copilot`, `gemini`, `claude`, `opencode`, `codex`, `auto` (default: `auto`). If multiple agents are enabled and user hasn't specified, the LLM must call `ask_user` first.
`workspace`	string	—	Target workspace (uses active workspace if not specified)
`timeout`	number	—	Task timeout in milliseconds (default: 300000 = 5 minutes)

Returns: Narrative text (e.g. "Created task tsk_Xy7z: 'Add error handling' → claude in my-project") with task data in metadata.

Danger Level: MODERATE

task_status

Check the status of a running task.

Parameters:

Name	Type	Required	Description
`taskId`	string	—	Task ID (returns current task if not specified)

Returns: Narrative text (e.g. "Task tsk_Xy7z running (45s) — 'Add error handling'. Last: installing deps") with full task state in metadata.

Danger Level: SAFE

task_cancel

Cancel a running task and terminate its active harness process when one is running.

Parameters:

Name	Type	Required	Description
`taskId`	string	✅	ID of the task to cancel

Returns: Narrative text (e.g. "Cancelled task tsk_Xy7z (was running 32s); terminated active process")

Danger Level: MODERATE

task_respond

Send user input to a task that is waiting for a response.

Parameters:

Name	Type	Required	Description
`response`	string	✅	Response to send to the waiting task
`taskId`	string	—	Task ID (uses current waiting task if not specified)

Returns: Narrative text (e.g. "Sent response to task tsk_Xy7z, resuming execution")

Danger Level: SAFE

Interaction Tools (2)

ask_user

Send a question to the user via WhatsApp and wait for a reply.

Parameters:

Name	Type	Required	Description
`question`	string	✅	Question to ask the user
`options`	string[]	—	Optional list of choices for the user (max 10 options, each max 100 chars)

Returns: { answer: string }

Autonomy Guard: In cautious or autonomous mode, questions matching unnecessary confirmation patterns ("Shall I...", "Would you like me to...", "Can I proceed...") are auto-approved without reaching the user. The LLM receives "Yes, proceed." as the answer. This is controlled by ToolContext.autonomyLevel.

Danger Level: SAFE

report_progress

Send a progress update to the user.

Parameters:

Name	Type	Required	Description
`message`	string	✅	Progress message to display
`percent`	number	—	Percentage complete (0-100)

Returns: { sent: boolean }

Danger Level: SAFE

GitHub Tools (8)

github_pr_create

Create a pull request on GitHub from the current branch.

Parameters:

Name	Type	Required	Description
`title`	string	✅	PR title (max 256 chars)
`body`	string	—	PR description (max 4000 chars)
`base`	string	—	Base branch to merge into (default: `main`)
`draft`	boolean	—	Create as a draft pull request (default: `true`)
`workspace`	string	—	Workspace (uses active workspace if not specified)

Returns: { url: string, number: number }

Danger Level: MODERATE

github_pr_list

List pull requests for the current or specified repository.

Parameters:

Name	Type	Required	Description
`state`	string	—	Filter by PR state: `open`, `closed`, `all` (default: `open`)
`repo`	string	—	Target repository in `org/repo` format (e.g. `facebook/react`)
`limit`	number	—	Maximum number of results (1-100, default: 10)
`workspace`	string	—	Workspace (uses active workspace if not specified)

Returns: object[] (array of PR objects)

Danger Level: SAFE

github_pr_view

View details of a specific pull request including reviews and comments.

Parameters:

Name	Type	Required	Description
`number`	number	✅	Pull request number
`repo`	string	—	Target repository in `org/repo` format
`workspace`	string	—	Workspace (uses active workspace if not specified)

Returns: object (PR details)

Danger Level: SAFE

github_issue_create

Create a new GitHub issue in the current repository.

Parameters:

Name	Type	Required	Description
`title`	string	✅	Issue title (max 256 chars)
`body`	string	—	Issue description (max 4000 chars)
`labels`	string[]	—	Labels to apply (max 10 labels, each max 50 chars)
`workspace`	string	—	Workspace (uses active workspace if not specified)

Returns: { url: string, number: number }

Danger Level: MODERATE

github_issue_list

List issues for the current repository with optional filters.

Parameters:

Name	Type	Required	Description
`state`	string	—	Filter by issue state: `open`, `closed`, `all` (default: `open`)
`assignee`	string	—	Filter by assignee username (max 39 chars)
`labels`	string[]	—	Filter by labels (max 10 labels)
`workspace`	string	—	Workspace (uses active workspace if not specified)

Returns: object[] (array of issue objects)

Danger Level: SAFE

github_branch_create

Create a new git branch and push it to GitHub.

Parameters:

Name	Type	Required	Description
`name`	string	✅	Branch name (max 100 chars, alphanumeric/hyphens/dots/slashes)
`from`	string	—	Branch to create from (defaults to current branch)
`workspace`	string	—	Workspace (uses active workspace if not specified)

Returns: object (branch details)

Danger Level: MODERATE

github_action_status

Get the status of recent GitHub Actions workflow runs.

Parameters:

Name	Type	Required	Description
`workspace`	string	—	Workspace (uses active workspace if not specified)

Returns: object[] (array of workflow run objects)

Danger Level: SAFE

github_search_repos

Search GitHub repositories by keyword.

Parameters:

Name	Type	Required	Description
`query`	string	✅	Search query (max 256 chars)
`limit`	number	—	Maximum number of results (1-20, default: 5)

Returns: object[] (array of repository objects)

Danger Level: SAFE

Web Tools (2)

web_fetch

Fetch a web page and extract its readable content as markdown. Uses jsdom + Mozilla Readability + Turndown. Blocks private/internal URLs.

Parameters:

Name	Type	Required	Description
`url`	string	✅	URL to fetch (must be a valid http/https URL)
`selector`	string	—	Optional CSS selector to extract specific content from the page (max 200 chars)

Returns: { url, title, content, truncated, length } (HTML pages) or { url, contentType, content } (JSON/plain text)

Security: Private/internal URLs (localhost, 127.x.x.x, 10.x.x.x, 172.16-31.x.x, 192.168.x.x, 0.0.0.0, ::1, fe80:) are blocked. Content is truncated at 50,000 characters. Requests time out after 30 seconds.

Danger Level: SAFE

web_search

Search the web using a self-hosted SearXNG meta search engine. Returns structured results with title, URL, snippet, and engine source.

Parameters:

Name	Type	Required	Description
`query`	string	✅	Search query (1-400 chars)
`count`	number	—	Number of results to return (1-20, default: 5)
`category`	string	—	Search category: `general`, `images`, `news`, `science`, `it` (default: `general`)

Returns: { query, count, results: [{ title, url, snippet, engine }] }

Requires: SearXNG container running (docker compose up -d). Searches time out after 15 seconds.

Danger Level: SAFE

Browser Tools (4)

browser_open

Open a URL in a headless browser and return an accessibility tree snapshot with numbered element refs.

Parameters:

Name	Type	Required	Description
`url`	string	✅	URL to navigate to (must be a valid URL)
`waitUntil`	string	—	Wait condition: `load`, `domcontentloaded`, or `networkidle` (default: `load`)

Returns: Accessibility tree snapshot text (from browser-agent.mjs stdout)

Danger Level: MODERATE

browser_snapshot

Get the current page's accessibility tree snapshot without navigating.

Parameters: none (empty object)

Returns: Accessibility tree snapshot text

Danger Level: SAFE

browser_action

Perform an action on the browser page using element ref numbers from the snapshot.

Parameters:

Name	Type	Required	Description
`action`	string	✅	Action type: `click`, `type`, `scroll_down`, `scroll_up`, `back`, `forward`
`ref`	number	—	Element ref number from browser_snapshot (required for `click` and `type`)
`text`	string	—	Text to type into the element (required for `type` action, max 2000 chars)
`x`	number	—	X coordinate for coordinate-based click
`y`	number	—	Y coordinate for coordinate-based click

Returns: Updated accessibility tree snapshot text

Scroll actions use scroll_down / scroll_up (not a separate direction parameter).

Danger Level: MODERATE

browser_screenshot

Capture a screenshot of the current browser page.

Parameters: none (empty object)

Returns: Base64-encoded PNG screenshot (from browser-agent.mjs stdout)

Danger Level: SAFE

Note: Browser tools require ENABLE_BROWSER=true and Playwright+Chromium installed in the Kennel container. Browser state persists across tool calls within a session. All browser commands execute inside fetch-kennel via docker exec with a 30-second timeout.

Workflow & Runtime Tools (11)

These tools are best understood in three layers:

Delegation: open-ended coding (task_*).
Interactive: exploratory research/browsing (web_*, browser session tools).
Execution: deterministic automation primitives (app_run, app_test, browser_test).

Workflow and cron automation are designed primarily for the Execution layer.

workflow_create / workflow_list / workflow_run / workflow_delete

Create, inspect, execute, and remove named workflows. A workflow is an ordered list of existing tool calls with optional workspace pre-selection. workflow_create validates step tools up-front and blocks recursive orchestration tools (workflow_*, cron_*) and task-only interaction tools (ask_user, report_progress).

cron_create / cron_list / cron_delete / cron_run

Manage UTC cron schedules for workflows. Schedules use 5-field cron format (minute hour day month weekday).

app_run / app_test

Execute deterministic commands inside the active (or specified) workspace within Kennel. app_test auto-detects a default command when possible (npm test, go test ./..., cargo test, pytest).

browser_test

Run a deterministic browser smoke check by opening a URL, collecting an accessibility snapshot, and asserting expected substrings (mustInclude), with optional screenshot capture.

Security Interfaces

SecurityGate

Controls message authorization.

Method	Description
`authorize(message)`	Returns `{ allowed: boolean, reason?: string }`
`isOwnerMessage(message)`	Checks if sender is `OWNER_PHONE_NUMBER`
`stripTrigger(text)`	Removes `@fetch` prefix

RateLimiter

Sliding window rate limiter.

Method	Description
`isAllowed(key)`	Returns `boolean` — checks if under rate limit
`remaining(key)`	Returns `number` — remaining requests in window
`clear(key)`	Reset a specific key's history
`clearAll()`	Reset all rate limit state

Default config: 30 requests per 60-second sliding window.

InputValidator

Validates and sanitizes user input.

Method	Description
`validate(input)`	Returns `{ valid: boolean, sanitized: string, reason?: string }`

Blocks: command substitution ($()), rm -rf patterns, pipe-to-shell, eval(), prototype pollution, null bytes, control characters. Allows backticks (for code discussion).

Tool Context

interface ToolContext {
  sessionId?: string;       // Session ID for session-aware tools
  autonomyLevel?: "supervised" | "cautious" | "autonomous"; // Safety/autonomy policy level
}

The autonomyLevel field flows from session preferences through the registry into tool execution.

ask_user uses it for confirmation auto-approval behavior.
The tool registry enforces dangerous-tool policy gates (supervised blocks dangerous tools; cautious requires explicit confirm: true).

Session Types

interface Session {
  id: string;                         // Session ID
  userId: string;                     // WhatsApp JID
  metadata: Record<string, any>;      // Flexible metadata blob
  messages: Message[];                // Conversation history
  currentProject: ProjectContext | null; // Active project
  availableProjects: string[];        // Known workspace names
  activeFiles: string[];              // Files in context
  repoMap: string | null;             // Cached repo map
  repoMapUpdatedAt: string | null;    // Repo map timestamp
  preferences: UserPreferences;       // Autonomy, verbose, autocommit
  activeTaskId: string | null;        // Currently running task
  gitStartCommit: string | null;      // Commit SHA for undo boundary
  createdAt: string;
  lastActivityAt: string;
}

Task Types

interface Task {
  id: string;
  sessionId: string;
  goal: string;
  status: TaskStatus;  // pending → running → completed | failed | cancelled
  harness: string;
  createdAt: string;
  updatedAt: string;
  result?: TaskResult;
  iterations: number;
  maxIterations: number;
}

Task status transitions: pending → running → waiting_input → running → completed | failed | cancelled

FilesExpand file tree

API_REFERENCE.md

Latest commit

History

API_REFERENCE.md

File metadata and controls

API Reference

Implementation References

Status API

Endpoint Summary

GET /api/status

GET /api/sessions

GET /api/sessions/:id

GET /api/health

POST /api/logout

POST /api/config/reload

DELETE /api/sessions/:id

POST /api/sessions/:id/clear

Orchestrator Tools

Tool Module Ownership

Supporting Module Ownership

Workspace Tools (9)

workspace_list

workspace_select

workspace_status

workspace_create

workspace_delete

file_delete

folder_delete

workspace_sync

workspace_publish

Task Tools (4)

task_create

task_status

task_cancel

task_respond

Interaction Tools (2)

ask_user

report_progress

GitHub Tools (8)

github_pr_create

github_pr_list

github_pr_view

github_issue_create

github_issue_list

github_branch_create

github_action_status

github_search_repos

Web Tools (2)

web_fetch

web_search

Browser Tools (4)

browser_open

browser_snapshot

browser_action

browser_screenshot

Workflow & Runtime Tools (11)

workflow_create / workflow_list / workflow_run / workflow_delete

cron_create / cron_list / cron_delete / cron_run

app_run / app_test

browser_test

Security Interfaces

SecurityGate

RateLimiter

InputValidator

Tool Context

Session Types

Task Types