Skip to main content
Glama
Jules MCP Server

Jules MCP Server

by cipansor

mcp-jules

CI npm version npm downloads node license

MCP (Model Context Protocol) server for Google Jules - the AI coding agent. This server enables LLMs like Claude to interact with Jules API, creating and managing coding sessions programmatically.

Unofficial project. This is an independent, community-built client for the Jules API. It is not affiliated with, sponsored by, or endorsed by Google LLC. "Google" and "Jules" are trademarks of Google LLC. The Jules API is an alpha release and its specifications may change without notice.

Table of Contents

Related MCP server: Jules MCP Server

Architecture

The Jules MCP Server acts as a bridge between MCP-compatible clients and the Jules REST API.

┌──────────────┐      ┌──────────────────┐      ┌──────────────────┐
│  MCP Client  │      │ Jules MCP Server │      │  Jules REST API  │
│ (Claude, etc)│ ◄──► │ (stdio/HTTP)     │ ◄──► │ (googleapis.com) │
└──────────────┘      └──────────────────┘      └──────────────────┘
  1. MCP Client: An application like Claude Desktop that supports the Model Context Protocol.

  2. Jules MCP Server: This project, which exposes Jules's capabilities as MCP tools and translates them to Jules API calls.

  3. Jules REST API: The underlying Google API that powers Jules's coding capabilities.

Features

  • Session Management: Create, list, get, and delete coding sessions

  • Interactive Communication: Send messages to active sessions and approve plans

  • Activity Monitoring: Track progress through session activities

  • Source Management: List and inspect connected GitHub repositories

  • Dual Transport: Supports both stdio and HTTP transports

  • Hardened HTTP Transport: Loopback-by-default bind, Host/Origin validation (DNS-rebinding protection), and optional bearer-token authentication

Prerequisites

Installation

mcp-jules is published on npm as mcp-jules and installs a mcp-jules command.

# Install globally
npm install -g mcp-jules

# Or run it without installing (recommended for MCP clients)
npx -y mcp-jules --help

The package requires Node.js 18 or newer and declares no post-install scripts. npx -y is the most common way to run it, because MCP clients start the server as a child process.

From source

git clone https://github.com/cipansor/mcp-jules
cd mcp-jules
npm install
npm run build
node dist/index.js --help

Quick Start

Three steps from zero to a working server:

  1. Get an API key. Create one at jules.google.com/settings and make sure at least one GitHub repository is connected to Jules in the Jules web UI.

  2. Export the key. The server refuses to start without it:

    export JULES_API_KEY="your-api-key-here"
  3. Register the server with your MCP client. For Claude Desktop, add the mcpServers entry shown in Usage.

Then ask your client to call jules_list_sources - if it returns your repositories, the wiring works.

Configuration

Environment variables

Variable

Required

Default

Description

JULES_API_KEY

Yes

-

Jules API key. The server exits with an error when it is missing.

TRANSPORT

No

stdio

Set to http to use the HTTP transport instead of stdio (same as --http).

HOST

No

127.0.0.1

HTTP bind address. Only used by the HTTP transport.

PORT

No

3000

HTTP port. Only used by the HTTP transport. Invalid values are rejected.

Set the key as an environment variable:

export JULES_API_KEY="your-api-key-here"

The key is sent to the Jules API in the x-goog-api-key header and is never written to disk by this server.

HTTP security variables

These apply only to the HTTP transport. See HTTP transport hardening for how they interact.

Variable

Default

Description

MCP_AUTH_TOKEN

unset

When set, /mcp requires Authorization: Bearer <token>. Mandatory for any non-loopback bind.

ALLOWED_HOSTS

unset

Comma-separated Host header allowlist. Required when binding 0.0.0.0 or ::.

ALLOWED_ORIGINS

loopback origins

Comma-separated Origin allowlist. Requests from unlisted origins get 403.

ALLOW_INSECURE_HTTP

unset

Set to true to permit an unauthenticated non-loopback bind. Not recommended.

Security

Manage your JULES_API_KEY with care:

  • Environment Variables: Always use environment variables to provide the API key. Avoid hardcoding it in source code or configuration files.

  • No Commits: Never commit your API key or .env files containing keys to version control. The .gitignore file is configured to ignore .env files.

  • Least Privilege: The Jules API key provides full access to your Jules sessions and connected repositories. Keep it secure and rotate it if you suspect it has been compromised.

  • CI/CD: When using in CI/CD environments (like GitHub Actions), use Secrets to store and inject the API key.

HTTP transport hardening

The HTTP transport can drive mutating Jules tools with your API key, so it is locked down by default:

  • Loopback by default: the server binds 127.0.0.1. It refuses to bind a non-loopback address unless MCP_AUTH_TOKEN is set and ALLOWED_HOSTS lists the hostnames clients will use. ALLOW_INSECURE_HTTP=true overrides this refusal and is not recommended.

  • DNS-rebinding protection: the Host header is validated against the allowed hostnames, and the MCP transport rejects requests from origins outside ALLOWED_ORIGINS.

  • Optional authentication: set MCP_AUTH_TOKEN to require Authorization: Bearer <token> on /mcp. /health stays unauthenticated.

For example, to expose the server on a trusted LAN:

JULES_API_KEY=your-key \
  HOST=0.0.0.0 \
  MCP_AUTH_TOKEN=a-long-random-token \
  ALLOWED_HOSTS=mcp.internal \
  ALLOWED_ORIGINS=https://mcp.internal \
  npx mcp-jules --http

ALLOWED_ORIGINS must include the origin each browser client sends. If your clients run on https://app.example.com, add it to the comma-separated list; requests from origins that are not listed receive 403.

If you deploy behind a reverse proxy, terminate TLS at the proxy and keep ALLOWED_ORIGINS restricted to the exact origins of your MCP clients.

Do not set ALLOWED_HOSTS to a name that resolves elsewhere while binding loopback without a token: a tokenless listener only accepts hostnames that resolve to the local machine.

See GHSA-rgh8-8cv7-4rj9 for the advisory that prompted these controls.

Usage

With Claude Desktop (stdio transport)

Add to your Claude Desktop configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "jules": {
      "command": "npx",
      "args": ["-y", "mcp-jules"],
      "env": {
        "JULES_API_KEY": "your-api-key-here"
      }
    }
  }
}

With Claude Code (CLI)

You can add this server to Claude Code by running:

claude mcp add jules --env JULES_API_KEY=your-api-key-here -- npx -y mcp-jules

Command Line

# Run with stdio transport
JULES_API_KEY=your-key npx mcp-jules

# Run with HTTP transport (binds 127.0.0.1 by default)
JULES_API_KEY=your-key npx mcp-jules --http

# Custom port for HTTP
JULES_API_KEY=your-key PORT=8080 npx mcp-jules --http

# Expose on a trusted LAN (token + explicit host allowlist required)
JULES_API_KEY=your-key HOST=0.0.0.0 MCP_AUTH_TOKEN=your-token \
  ALLOWED_HOSTS=mcp.internal ALLOWED_ORIGINS=https://mcp.internal \
  npx mcp-jules --http

# Show help
npx mcp-jules --help

See HTTP transport hardening for the full list of security-related environment variables.

HTTP Endpoints

The HTTP transport exposes exactly two routes:

Route

Method

Auth

Description

/mcp

POST

Bearer token if MCP_AUTH_TOKEN is set

MCP streamable-HTTP endpoint. This is what MCP clients connect to.

/health

GET

none

Liveness probe. Returns {"status":"ok","server":"mcp-jules","version":"..."}.

The version in /health is read from package.json at runtime, so it always matches the published npm version. /health is intentionally unauthenticated so a container orchestrator can probe it without a token; it exposes no session data.

# Start the HTTP transport
JULES_API_KEY=your-key npx mcp-jules --http

# Probe liveness
curl http://127.0.0.1:3000/health

Tool Reference

The server exposes ten tools across three groups. Inputs are validated with a strict schema, so an unknown argument is rejected rather than ignored.

Tool

Group

response_format

Paginated

jules_create_session

Sessions

yes

no

jules_list_sessions

Sessions

yes

yes

jules_get_session

Sessions

yes

no

jules_delete_session

Sessions

no

no

jules_send_message

Sessions

no

no

jules_approve_plan

Sessions

no

no

jules_list_activities

Activities

yes

yes

jules_get_activity

Activities

yes

no

jules_list_sources

Sources

yes

yes

jules_get_source

Sources

yes

no

Every tool that accepts response_format defaults to markdown and also accepts json. Paginated tools accept pageSize (1-100, default 20) and pageToken, and report whether more results are available. In markdown output, jules_get_activity truncates long artifact bodies — git patches at 5,000 characters and bash output at 2,000 characters — appending ... [truncated]. JSON output is never truncated.

Session Tools

jules_create_session

Creates a new coding session where Jules executes a task on a repository.

  • Arguments:

    • prompt (string, required): Task description. Max 10,000 characters.

    • sourceContext (object, required): Repository info (e.g., { "source": "sources/github-owner-repo", "githubRepoContext": { "startingBranch": "main" } }). githubRepoContext is optional.

    • title (string, optional): Session title. Max 200 characters.

    • requirePlanApproval (boolean, optional): If Jules should wait for plan approval.

    • automationMode (string, optional): Use AUTO_CREATE_PR to automatically create a PR on completion.

    • response_format (string, optional): Response format, either markdown (default) or json.

  • Example Response (Markdown):

    ## Session: Add auth tests
    
    **ID:** 1234567
    **State:** ⏳ QUEUED
    **Prompt:** Add unit tests for the authentication module
    **URL:** https://jules.google/session/1234567

jules_list_sessions

Lists all your coding sessions.

  • Arguments:

    • pageSize (number, optional): Max results per page (1-100, default: 20).

    • pageToken (string, optional): Token for the next page.

    • response_format (string, optional): Format for the response output.

  • Example Response (JSON):

    {
      "sessions": [
        {
          "id": "1234567",
          "title": "Add auth tests",
          "state": "IN_PROGRESS",
          "createTime": "2023-10-27T10:00:00Z"
        }
      ],
      "total": 1,
      "hasMore": false
    }

jules_get_session

Retrieves full details for a specific session, including current state, URL, and outputs (like PR links) if completed.

  • Arguments:

    • sessionId (string, required): The ID of the session.

    • response_format (string, optional): 'markdown' (default) or 'json'.

  • Example Response (Markdown):

    ## Session: Add auth tests
    
    **ID:** 1234567
    **State:** ✅ COMPLETED
    **Prompt:** Add unit tests for the authentication module
    **URL:** https://jules.google/session/1234567
    
    **Created:** Oct 27, 2023, 10:00 AM
    **Updated:** Oct 27, 2023, 10:45 AM
    
    ### Outputs
    
    **Pull Request:** [Add auth unit tests](https://github.com/myorg/myrepo/pull/42)
    
    > This PR adds comprehensive unit tests for the auth module.
  • Example Response (JSON):

    {
      "id": "1234567",
      "title": "Add auth tests",
      "prompt": "Add unit tests for the authentication module",
      "state": "COMPLETED",
      "url": "https://jules.google/session/1234567",
      "outputs": [
        {
          "pullRequest": {
            "title": "Add auth unit tests",
            "url": "https://github.com/myorg/myrepo/pull/42",
            "description": "This PR adds comprehensive unit tests for the auth module."
          }
        }
      ],
      "createTime": "2023-10-27T10:00:00Z",
      "updateTime": "2023-10-27T10:45:00Z"
    }

jules_delete_session

Deletes a coding session. This cannot be undone.

  • Arguments:

    • sessionId (string, required): The ID of the session.

jules_send_message

Sends a message to an active session. Use this to provide feedback, answer Jules's questions, or change instructions mid-session.

  • Arguments:

    • sessionId (string, required): The ID of the session.

    • prompt (string, required): Your message.

jules_approve_plan

Approves a pending plan. Required if requirePlanApproval was set to true during session creation. Once approved, Jules begins execution.

  • Arguments:

    • sessionId (string, required): The ID of the session.

Activity Tools

jules_list_activities

Lists all activities (events, plan generation, code changes, messages) for a session.

  • Arguments:

    • sessionId (string, required): The ID of the session.

    • pageSize (number, optional): Results per page. Must be between 1 and 100. Defaults to the schema-defined default when omitted.

    • pageToken (string, optional): Pagination token.

    • response_format (string, optional): Response format for the returned activities.

jules_get_activity

Gets detailed information about a specific activity. This is how you view code diffs (changeSet), bash outputs, or full plan details.

  • Arguments:

    • sessionId (string, required): The ID of the session.

    • activityId (string, required): The activity ID.

    • response_format (string, optional): Controls the format of the returned activity details.

  • Note: in markdown output, a git patch is truncated to 5,000 characters and bash output to 2,000 characters, each ending with ... [truncated]. Use response_format: "json" to receive the full, untruncated payload.

Source Tools

jules_list_sources

Lists GitHub repositories connected to your Jules account via the Jules web UI.

  • Arguments:

    • filter (string, optional): Filter by name (e.g., name=sources/github-owner-repo).

    • pageSize (number, optional): Results per page.

    • pageToken (string, optional): Pagination token.

    • response_format (string, optional): Response format for the returned data.

jules_get_source

Gets details about a specific connected repository, including all available branches.

  • Arguments:

    • sourceId (string, required): The source ID (e.g., github-owner-repo). This is not the sources/... resource name used by sourceContext.source.

    • response_format (string, optional): Controls the response format returned by the tool.

Real-World Examples

Complete Workflow: Implementing a Feature

  1. Find the repository ID: jules_list_sources() -> returns sources/github-acme-webapp.

  2. Start the session: jules_create_session(prompt="Add a search bar to the header", sourceContext={"source": "sources/github-acme-webapp"}, requirePlanApproval=true) -> returns sessionId 7890.

  3. Check for a plan: jules_list_activities(sessionId="7890"). Look for an activity with "Plan Generated".

  4. Review and approve: jules_get_activity(sessionId="7890", activityId="plan_activity_id"). If the plan looks good: jules_approve_plan(sessionId="7890").

  5. Monitor progress: Periodically call jules_get_session(sessionId="7890") to check state or jules_list_activities to see recent actions.

  6. Review the result: Once state is COMPLETED, check jules_get_session for the Pull Request URL.

Session States

State

Description

QUEUED

Session created, waiting for processing.

PLANNING

Jules is analyzing the codebase and creating a plan.

AWAITING_PLAN_APPROVAL

Plan is ready and requires your approval to proceed.

AWAITING_USER_FEEDBACK

Jules has a question or needs more information.

IN_PROGRESS

Jules is executing the task/plan.

PAUSED

Execution has been temporarily halted.

COMPLETED

Task finished successfully (check for PR link).

FAILED

Task failed. Check activities for error details.

Troubleshooting

Error / Symptom

Meaning

Resolution

Error: JULES_API_KEY environment variable is required.

Missing API key

Export JULES_API_KEY before starting the server. The process exits immediately without it.

401 Unauthorized

Invalid API Key

Check your JULES_API_KEY environment variable and ensure it's valid.

403 Forbidden

Permission Denied

Ensure your API key has access to the requested resource or repository.

404 Not Found

Resource Missing

Verify the sessionId, activityId, or sourceId is correct.

429 Too Many Requests

Rate Limited

You've exceeded the API rate limit. Wait a few minutes before retrying.

500 / 503

Jules API Error

The Jules service is experiencing issues. Try again later or check Jules Status.

Network error - unable to reach Jules API

DNS / connectivity

Check your internet connection, DNS, or proxy settings.

Network error: ...

Transport failure

Usually a timeout (ECONNABORTED) or a refused connection. Retry, then check firewall rules.

... [truncated] at the end of output

A markdown artifact body exceeded its limit (5,000 chars for a git patch, 2,000 for bash output)

Re-run jules_get_activity with response_format: "json" for the untruncated payload.

403 from the HTTP transport

Origin not allowed

Add the client origin to ALLOWED_ORIGINS.

Refuses to bind a non-loopback HOST

Hardening guardrail

Set both MCP_AUTH_TOKEN and ALLOWED_HOSTS before binding 0.0.0.0.

Server exits with an invalid PORT

Bad configuration

PORT must be an integer between 1 and 65535.

Errors are returned to the MCP client as tool errors; the original Jules API status code is preserved in the message as Jules API Error (<status>): ....

Development

# Install dependencies
npm install

# Build (emits dist/ from src/)
npm run build

# Type-check without emitting files
npm run check

# Lint source files
npm run lint

# Check formatting
npm run format:check

# Run the test suite (Node's built-in test runner)
npm test

# Development mode (watch)
npm run dev

The build compiles TypeScript to dist/, which is what the mcp-jules bin and the npm tarball ship. dist/ is generated, so edit files under src/ only. The CI workflow runs the format check, lint, type-check and tests on every push and pull request.

Contributing

We welcome contributions! See CONTRIBUTING.md for more details.

  1. Check Issues: Look for existing issues or open a new one to discuss your proposed change.

  2. Local Setup: Follow the Installation steps.

  3. Topic Branches: Always work on a new branch (git checkout -b feature/my-feature).

  4. Testing: If adding a tool, ensure it's tested. Run npm run check and npm run lint.

  5. Documentation: Update the README if you change tool behavior or add new features.

  6. Pull Requests: Submit a PR with a clear description of the changes.

Project hygiene and community documents:

Releasing

Releases are driven by a git tag. The tag, the package.json version and the npm version must all agree, so the version is bumped first and the tag is pushed at the merge commit:

  1. Bump the version: npm version <major.minor.patch> --no-git-tag-version.

  2. Merge to main, then tag the merge commit v<version> and push the tag.

Pushing a v*.*.* tag runs the release workflow, which verifies the tag matches package.json, builds, attaches the tarball to a GitHub release, and publishes to npm with provenance via Trusted Publishing.

One-time npm setup

Publishing uses Trusted Publishing, so the release workflow authenticates with a short-lived OIDC token and no NPM_TOKEN secret is stored in the repository. Trusted publishing can only be attached to a package that already exists in the registry, so the package was created once with a staged publish before the first tag-driven release.

1. Bootstrap the package (already done for mcp-jules). A brand-new package is created with staged publishing, which needs npm CLI >= 11.15.0 and Node >= 22.14.0:

npm stage publish --access public   # uploads the tarball to the stage queue
npm stage list                      # wait for status: staged, then copy the id
npm stage approve <stage-id>        # needs 2FA

Staging a package that does not exist yet creates a 0.0.0-stage placeholder so the name is reserved and the trusted publisher can be attached; the real version goes live only after npm stage approve. Staging itself does not need 2FA, which is why this path works from a token that cannot bypass 2FA - the proof-of-presence is deferred to the interactive approve step. A direct npm publish instead fails with 403 ... Two-factor authentication or granular access token with bypass 2fa enabled is required to publish packages.

Bootstrap with a throwaway version (for example 0.0.1) and leave the release you want CI to ship unpublished: a published version cannot be published again, so bootstrapping 0.2.0 would make the v0.2.0 tag fail in the publish job.

2. Add the trusted publisher. On npmjs.com, open the mcp-jules package, then Settings → Trusted Publisher → GitHub Actions and enter:

  • Organization or user: cipansor

  • Repository: mcp-jules

  • Workflow filename: release.yml (the filename only, not the full path)

  • Environment name: leave blank

  • Allowed actions: enable direct npm publish, not stage-only. The release workflow runs npm publish, so a stage-only connection makes the publish job fail.

The identity fields are exact-match, so a wrong repository or workflow filename is the usual cause of an ENEEDAUTH or 404 failure in the publish job. These fields cannot be edited after the connection is created; delete and recreate the connection to change them.

While you are there, set Publishing access to Require two-factor authentication and disallow bypass 2fa tokens. Trusted publishers keep working under that setting, because CI authenticates with OIDC rather than a long-lived token.

API Reference

Based on Jules REST API:

License

Apache-2.0. See LICENSE and NOTICE.

Copyright 2026 Yayasan Pesantren Cipansor and the mcp-jules contributors.

This is an independent project, not affiliated with, sponsored by, or endorsed by Google LLC. "Google" and "Jules" are trademarks of Google LLC, used here only to describe what this software interoperates with.

Available Tools

10 tools
jules_approve_planApprove Jules Session PlanA
Idempotent

Approve a pending plan in a Jules session.

Only needed when the session was created with requirePlanApproval=true. After approval, Jules will execute the planned steps.

Args:

  • sessionId (string, required): The session ID with pending plan

Returns: Confirmation that the plan was approved.

Note: Check session state is 'AWAITING_PLAN_APPROVAL' before calling.

Inputs: sessionId (required).

jules_create_sessionCreate Jules SessionA

Create a new coding session with Jules AI coding agent.

A session represents a unit of work where Jules executes a coding task on your repository.

Args:

  • prompt (string, required): The task description for Jules to execute

  • title (string, optional): Title for the session

  • sourceContext (object, required): Repository context with:

    • source (string): Resource name like 'sources/github-owner-repo'

    • githubRepoContext (object, optional): { startingBranch: string }

  • requirePlanApproval (boolean, optional): If true, plans need approval

  • automationMode (string, optional): 'AUTO_CREATE_PR' to auto-create PRs

  • response_format ('markdown' | 'json'): Output format

Returns: The created session with ID, state, and URL.

Examples:

  • "Add unit tests for auth module" with source "sources/github-myorg-myrepo"

  • Set requirePlanApproval=true to review plans before execution

Inputs: prompt (required), sourceContext (required), title, automationMode, +2 more.

jules_delete_sessionDelete Jules SessionA
DestructiveIdempotent

Delete a coding session.

Args:

  • sessionId (string, required): The session ID to delete

Returns: Confirmation of deletion.

Note: This action cannot be undone.

Inputs: sessionId (required).

jules_get_activityGet Session ActivityA
Read-onlyIdempotent

Get details of a specific activity in a Jules session.

Retrieves full details including artifacts like code changes, bash output, or media files.

Args:

  • sessionId (string, required): The session ID

  • activityId (string, required): The activity ID

  • response_format ('markdown' | 'json'): Output format

Returns: Full activity details including:

  • Event type (plan generated, message, progress, etc.)

  • Artifacts (code changes, command output, media)

  • Timestamps

Examples:

  • Get plan details: activity with planGenerated

  • Get code diff: activity with changeSet artifact

Inputs: sessionId (required), activityId (required), response_format.

jules_get_sessionGet Jules SessionA
Read-onlyIdempotent

Get details of a specific coding session.

Args:

  • sessionId (string, required): The session ID

  • response_format ('markdown' | 'json'): Output format

Returns: Full session details including:

  • State and progress

  • Source context

  • Outputs (pull requests if completed)

Examples:

  • Get session details: sessionId="1234567"

Inputs: sessionId (required), response_format.

jules_get_sourceGet Source DetailsA
Read-onlyIdempotent

Get details of a specific repository source.

Retrieves full information about a connected GitHub repository including all available branches.

Args:

  • sourceId (string, required): The source ID (e.g., 'github-owner-repo')

  • response_format ('markdown' | 'json'): Output format

Returns: Full source details:

  • Repository owner and name

  • Public/private status

  • Default branch

  • All available branches

Examples:

  • Get repo details: sourceId="github-myorg-myrepo"

Inputs: sourceId (required), response_format.

jules_list_activitiesList Session ActivitiesA
Read-onlyIdempotent

List all activities for a Jules coding session.

Activities track everything that happens during a session:

  • Plan generation and approval

  • User and agent messages

  • Progress updates

  • Completion or failure events

  • Code changes and artifacts

Args:

  • sessionId (string, required): The session ID

  • pageSize (number, optional): Results per page (1-100, default: 20)

  • pageToken (string, optional): Token for pagination

  • response_format ('markdown' | 'json'): Output format

Returns: List of activities with type, description, and timestamps.

Examples:

  • Monitor session progress: provide sessionId

  • Get full history: paginate through all activities

Inputs: sessionId (required), pageSize, pageToken, response_format.

jules_list_sessionsList Jules SessionsA
Read-onlyIdempotent

List all coding sessions for the authenticated user.

Args:

  • pageSize (number, optional): Results per page (1-100, default: 20)

  • pageToken (string, optional): Token for pagination

  • response_format ('markdown' | 'json'): Output format

Returns: List of sessions with ID, title, state, and timestamps. Includes nextPageToken if more results exist.

Examples:

  • List recent sessions: no params needed

  • Paginate: use pageToken from previous response

Inputs: pageSize, pageToken, response_format.

jules_list_sourcesList Connected SourcesA
Read-onlyIdempotent

List all repository sources connected to Jules.

Sources are GitHub repositories that Jules can work with. Sources are connected through the Jules web interface.

Args:

  • pageSize (number, optional): Results per page (1-100, default: 20)

  • pageToken (string, optional): Token for pagination

  • filter (string, optional): Filter expression (e.g., 'name=sources/github-owner-repo')

  • response_format ('markdown' | 'json'): Output format

Returns: List of sources with repository details:

  • Owner and repo name

  • Public/private status

  • Available branches

Examples:

  • List all repos: no params needed

  • Filter by name: filter="name=sources/github-myorg-myrepo"

Inputs: filter, pageSize, pageToken, response_format.

jules_send_messageSend Message to Jules SessionA

Send a message to an active Jules session.

Use this to provide feedback, answer questions, or give additional instructions while Jules is working on a task.

Args:

  • sessionId (string, required): The session ID

  • prompt (string, required): The message to send

Returns: Confirmation that the message was sent.

Examples:

  • "Please also add integration tests"

  • "Use TypeScript instead of JavaScript"

  • Answer a clarifying question from Jules

Inputs: prompt (required), sessionId (required).

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.3/5.0

    Scored across 10 tools

    Disambiguation5/5

    Each tool targets a distinct resource-action pair: sessions (create/get/list/delete/send/approve), activities (list/get), and sources (list/get). No two tools overlap in purpose, and descriptions clearly indicate which resource each operates on.

    Naming Consistency5/5

    All tool names follow the consistent jules_<verb>_<noun> pattern in snake_case, with verbs like get, list, create, delete, send, and approve. This makes the API highly predictable and easy to navigate.

    Tool Count5/5

    With 10 tools, the server is well-scoped for its purpose of managing AI coding sessions. Each tool has a clear, non-redundant role, covering session lifecycle, activity monitoring, and source context without unnecessary bloat.

    Completeness4/5

    The session lifecycle is well covered: create, read, list, delete, message, and plan approval. Activity and source listing/retrieval support monitoring and context. The only notable gap is a dedicated cancel/stop session tool, though sending a message could serve that purpose.

    Maintenance

    ActivityNo data
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers

    • -
      license
      A
      quality
      Not graded
      maintenance
      Enables automation of Google Jules AI coding assistant through task creation, code review automation, repository management, and AI-powered development workflows. Supports multiple session modes including cloud deployment with persistent authentication.
      13
      -
    • F
      license
      A
      quality
      D
      maintenance
      Enables LLM applications to interact with Google's Jules AI coding assistant to manage repositories, coding sessions, and pull requests. It allows users to programmatically create tasks, approve plans, and communicate with the assistant during active coding sessions.
      9
      -
    • A
      license
      B
      quality
      Not graded
      maintenance
      Exposes Google Jules AI capabilities for automated coding tasks, including session management, code reviews, and unified diff handling. It enables users to create sessions, approve plans, and synchronize AI-generated code changes with GitHub repositories.
      26
      10 npm
      MIT