Google Jules MCP Server
This MCP server lets local agents delegate and manage long-running coding tasks on Google Jules over stdio.
List configured Jules/GitHub accounts and routing aliases.
List connected GitHub repositories/sources and their default branches.
Delegate autonomous coding, refactoring, bug-fixing, or documentation tasks to Jules, optionally specifying target files, test command, base branch, PR creation, and plan approval.
Retrieve session status, details, outputs, PR links, and metadata.
List recent active and completed sessions.
Inspect a session's activity timeline, including plans, bash outputs, diffs, and PRs.
Approve a paused session plan so execution can proceed.
Send follow-up guidance, feedback, or adjustments to an ongoing session.
Extract the unified git diff/patch from a session.
Sources, status, sessions, activities, and diffs are read-only; delegation creates a remote session, plan approval starts execution, and messaging affects the remote session.
It does not run Jules locally, auto-merge PRs, or maintain background watcher loops.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Google Jules MCP Server List the repositories Jules can access."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
module_type: mcp-server status: active protocol: stdio primary_capability: Task delegation to Google Jules requires: JULES_API_KEY works_with: GitHub repositories, Google Jules last_verified: 2026-09-11
Google Jules MCP Server (Go)
Delegate complex, long-running coding tasks from local agents to the cloud-based Google Jules AI agent.
📚 Documentation Suite: Architecture • Agent Directives • Contributing • Security Policy • Specification • License
Related MCP server: Jules API MCP
Status and last verified date
Status: Active
Language: Go (Golang)
Protocol: Model Context Protocol (stdio transport)
Last verified: 2026-09-11
What it does / does not do
What it does:
Exposes Google Jules v1alpha REST API as a high-performance Model Context Protocol (stdio) server.
Lists connected GitHub repositories with their resolved default branches.
Autonomously delegates tasks to Jules, creating remote sessions in Google Cloud VMs.
Provides complete lifecycle observability: status checks, activity logs with unidiff patch previews, interactive message replies, and human plan approvals.
Implements smart branch resolution (R1): resolves default branch automatically from source metadata without hardcoded fallbacks.
Compact response formatting (R3): protects LLM context windows by truncating patch previews (<=300 chars) and prompt echoes (<=200 chars).
What it does not do:
Does not run Jules locally (Jules runs inside Google Cloud VMs).
Does not automatically approve or merge Jules's Pull Requests on GitHub.
Does not maintain unmonitored background watcher loops (one-shot actions only).
Quick start
Option A: Install via go install (Recommended)
go install github.com/TheNovaNodes/google-jules-mcp/cmd/google-jules-mcp@latestOption B: Build from source
# 1. Clone repository
git clone https://github.com/TheNovaNodes/google-jules-mcp.git
cd google-jules-mcp
# 2. Configure environment
cp .env.example .env
# Edit .env and set your JULES_API_KEY
# 3. Build binary
make build
# or: go build -ldflags="-s -w" -o bin/google-jules-mcp ./cmd/google-jules-mcp
# 4. Run server (stdio)
./bin/google-jules-mcpOption C: Production Deployment (5-Echelon Automated Gate)
In accordance with TheNovaNodes ecosystem standards, production deployment is orchestrated via ./scripts/deploy.sh:
# Execute standard 5-echelon deployment
./scripts/deploy.sh
# Dry-run validation (build and test without binary swap)
./scripts/deploy.sh --dry-runDeployment Echelons:
🛡️ Échelon 0: Pre-Flight Guard & Rollback Capsule: Toolchain checks, resolves
TARGET_COMMIT, and archives working binary into.deploy_rollback_capsule/.🧱 Échelon 1: Production Isolation Probe: Executes full test suite (
go test ./...) and ensures git working tree cleanliness.🔄 Échelon 2: Atomic Switch & Graceful Shift: Atomically installs candidate binary into
bin/google-jules-mcp.🪞 Échelon 3: Identity & Truth Probe: Queries
--versionto verify runtime commit SHA matchesgit rev-parse HEAD.🔍 Échelon 4: Deep Smoke Probe & Rollback Gate: Issues MCP stdio
initializehandshake probe. Automatic instant rollback (trap rollback ERR EXIT) on any anomaly.
Configuration and environment variables
JULES_API_KEY: The API key for accessing Google Jules REST API (Required).JULES_DISABLE_RETRY: Set to1ortrueto disable exponential backoff retries (Optional, for deterministic testing).
Complete MCP tool/API table with side effects
Tool | Description | Side Effects |
| List all GitHub repositories and default branches connected to Jules | None (Read-only) |
| Delegate a long-running coding/refactoring mission to Jules | Creates remote cloud session |
| Check the current execution state and status of a Jules session | None (Read-only) |
| Retrieve comprehensive details and metadata of a Jules session | None (Read-only) |
| List execution activity events, progress steps, and code patch previews | None (Read-only) |
| Send an interactive message or instructions to a Jules session | Sends message to remote session |
| Approve a proposed execution plan for a session paused in plan approval | Releases human gate; starts execution |
| Extract the latest Git unidiff patch produced by Jules for a session | None (Read-only) |
Security model and trust boundaries
Authentication: All upstream requests require a valid
JULES_API_KEYsent via theX-Goog-Api-Keyheader.Guardrails (R5): The server implements no background auto-approvers or watchers.
approve_jules_planis an explicit, one-shot action that logs a mandatory warning.Git Safety: The server never pushes directly, merges, or touches branches outside of what the Jules API executes remotely.
Generic MCP-client example
Add to your Claude Desktop, Antigravity, Cursor, or OpenClaw MCP configuration:
{
"mcpServers": {
"google-jules": {
"command": "/absolute/path/to/google-jules-mcp/bin/google-jules-mcp",
"args": [],
"env": {
"JULES_API_KEY": "YOUR_JULES_API_KEY"
}
}
}
}Testing
make test
# or: go test -v -race ./...License
MIT
Available Tools
9 toolsjules_approve_planA
Approve the generated execution plan for a session paused in AWAITING_PLAN_APPROVAL state.
Inputs: session_id (required), account_id.
jules_delegate_taskA
Delegate an autonomous coding, refactoring, bug-fixing, or documentation task to Google Jules. Jules will spin up a Google Cloud VM sandbox, clone the repo, plan changes, run test commands, apply modifications, and open a Pull Request automatically. Includes prompt boosting & verification contracts.
Inputs: prompt (required), source_name (required), account_id, target_files, +4 more.
jules_get_diffA
Extract the unified git patch diff from a completed or active Jules session.
Inputs: session_id (required), account_id.
jules_get_sessionA
Retrieve status, generated outputs, Pull Request links, and metadata for a specific Jules session.
Inputs: session_id (required), account_id.
jules_list_accountsA
List all configured Google Jules / GitHub accounts and routing aliases in the system.
Inputs: no named parameters.
jules_list_activitiesA
Get the chronological event timeline of a session (agent thoughts, execution plans, bash outputs, diffs, PRs).
Inputs: session_id (required), account_id.
jules_list_sessionsA
List recent active and completed Jules sessions for an account.
Inputs: page_size, account_id.
jules_list_sourcesA
List all GitHub repositories connected to Google Jules for a specific account or across defaults.
Inputs: page_size, account_id.
jules_send_messageB
Send follow-up guidance, feedback, or adjustments to an ongoing Jules session.
Inputs: message (required), session_id (required), account_id.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 9 tools
Each tool targets a distinct resource or action: accounts, sources, task delegation, session retrieval, session listing, activity timelines, plan approval, messaging, and diff extraction. There is minimal overlap, and any possible confusion between get_session, list_sessions, and list_activities is resolved by their distinct purposes.
All tools follow the consistent 'jules_verb_noun' pattern with snake_case, making the naming predictable and easy to navigate. For example, list_accounts, delegate_task, get_session, and approve_plan all use the same verb-first convention.
With 9 tools, the set is well-scoped for managing Google Jules sessions, covering account/source introspection, task delegation, monitoring, and interaction. The count is within the ideal 3-15 range and each tool earns its place.
The lifecycle covers delegation, monitoring (status, activities, diffs), approval, and follow-up messaging. A minor gap is the absence of a cancel/stop session operation, but the existing surface handles the primary workflows effectively.
Maintenance
Related MCP Connectors
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
MCP server for secureFlows: token-free URL builders and integration-linting tools for AI agents.
The Remote MCP server acts as a standardized bridge between LLM applications (like Claude, ChatGPT, and Cursor) and external services, enabling AI agents to access external tools and resources. Its primary capability is providing a centralized search tool to discover other MCP servers and their respective tools. Unlike local implementations, it runs remotely with OAuth authentication and permission controls for security.
Related MCP Servers
- AlicenseBqualityDmaintenanceAn MCP server for automating Google Jules that enables seamless integration for task creation, code analysis, and AI-powered development workflows. It supports multiple session modes including Browserbase and cookie-based authentication for both local and cloud environments.1314MIT
- AlicenseNot gradedqualityDmaintenanceAn MCP server that enables users to manage Google's Jules AI coding agent sessions directly from MCP-compatible clients. It supports creating sessions, approving execution plans, and interacting with session activity to streamline autonomous coding workflows.MIT
- AlicenseAqualityDmaintenanceAn MCP server for orchestrating Google Jules as a remote coding agent from a local coding agent, handling task decomposition, API dispatch, monitoring, intervention, code review, and PR merging.166 npmMIT
- AlicenseAqualityDmaintenanceMCP server for Google Jules enabling LLMs to create coding sessions with automatic pull request creation from issues or custom prompts.4MIT