devin-mcp
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., "@devin-mcp Create a Devin session to refactor the login module"
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.
devin-mcp
MCP server for creating, monitoring, and managing Devin AI sessions.
Tools
delegate
Create a Devin session and monitor it until completion. Runs as a background task with live progress updates (status changes, messages).
Supports all Devin session options:
Parameter | Description |
| The instruction for Devin to execute (required) |
| Custom session name (auto-generated if omitted) |
| Restore from a previous snapshot |
| Associated playbook identifier |
| Session categorization labels |
| Resource consumption ceiling |
| Prevent duplicate sessions with the same prompt |
| Hide session from listings |
| Knowledge bases to include ( |
| Secrets to include ( |
get_session
Retrieve details about an existing Devin session, including its status, messages, and metadata.
Parameter | Description |
| The identifier of the session to retrieve (required) |
list_sessions
List Devin sessions with optional filtering. Useful for finding session IDs to inspect or resume.
Parameter | Description |
| Maximum number of sessions to return (default 100) |
| Pagination offset (default 0) |
| Filter sessions by tags |
| Filter sessions by creator's email |
resume_session
Send a message to an existing Devin session and monitor it until completion. Runs as a background task. Use this to wake a sleeping session or send follow-up instructions to a running one.
Parameter | Description |
| The identifier of the session to message (required) |
| The message to send (required) |
Related MCP server: AgentHub
Requirements
Python 3.13+
Devin API key (starts with
apk_)
Usage
Claude Code
claude mcp add devin -e DEVIN_API_KEY=apk_your_key_here -- uvx --from git+https://github.com/desertaxle/devin-mcp devin-mcpStandalone
Run the MCP server directly:
uvx --from git+https://github.com/desertaxle/devin-mcp devin-mcpDevelopment
Install dev dependencies:
uv syncRun tests:
uv run pytestRun linter, formatter, and type checker:
uv run prek run --all-filesAvailable Tools
4 toolsdelegateA
Delegate a task to Devin and monitor until completion.
Creates a new Devin session with the given prompt and monitors it until the session reaches a terminal state (finished, blocked, or expired). Progress updates are reported as the session executes.
Args: prompt: The instruction for Devin to execute. title: Custom session name. Auto-generated if not provided. snapshot_id: Restore from a previous snapshot. playbook_id: Associated playbook identifier. tags: Session categorization labels. max_acu_limit: Resource consumption ceiling (positive integer). idempotent: If true, prevents duplicate sessions with same prompt. unlisted: If true, hides session from listings. knowledge_ids: Knowledge bases to include. None uses all, empty list uses none. secret_ids: Secrets to include. None uses all, empty list uses none. progress: FastMCP Progress dependency for reporting status updates.
Returns: Final session details including status, messages, and metadata.
Inputs: prompt (required), tags, title, unlisted, +6 more. Returns: object
get_sessionA
Retrieve details about an existing Devin session.
Use this to inspect the current status, messages, and metadata of a session. This is useful for checking whether a session is still running, has finished, or has gone to sleep due to ACU limits.
Args: session_id: The identifier of the session to retrieve.
Returns: Session details including status_enum, messages, title, tags, and metadata. The status_enum field indicates the session state: working, blocked, expired, finished, suspend_requested, resume_requested, or resumed.
Inputs: session_id (required). Returns: object
list_sessionsA
List Devin sessions with optional filtering.
Use this to find sessions by tags or email, or to get an overview of recent sessions. Useful for finding session IDs to inspect or resume.
Args: limit: Maximum number of sessions to return (default 100). offset: Pagination offset (default 0). tags: Filter sessions by these tags. user_email: Filter sessions by the creator's email address.
Returns: A dict with a 'sessions' key containing a list of session summaries, each with session_id, status_enum, title, tags, and other metadata.
Inputs: tags, limit, offset, user_email. Returns: object
resume_sessionA
Send a message to a Devin session and monitor it until completion.
Use this to resume a session that has gone to sleep (e.g. due to ACU limits) or to send follow-up instructions to a running session. Sending a message to a sleeping session will wake it up. After sending the message, the session is monitored until it reaches a terminal state.
Args: session_id: The identifier of the session to message. message: The message to send (e.g. instructions to continue work). progress: FastMCP Progress dependency for reporting status updates.
Returns: Final session details including status, messages, and metadata.
Inputs: message (required), session_id (required). Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 4 tools
Each tool has a clearly distinct purpose: get_session inspects a single session, list_sessions finds sessions, delegate creates and runs a new task, and resume_session continues an existing one. No two tools overlap in a way that would cause misselection.
Three tools follow the verb_noun pattern (get_session, list_sessions, resume_session), but 'delegate' is a bare verb without a noun, deviating slightly from the otherwise consistent style. The names are still predictable and readable.
Four tools is well-scoped for a Devin session management server. Each tool covers a distinct lifecycle action without redundancy, fitting comfortably within the ideal 3-15 range.
The set covers create (delegate), read (get_session, list_sessions), and update via message (resume_session), but lacks a delete/cancel operation for sessions. This is a minor gap that agents could work around.
Maintenance
Related MCP Connectors
MCP server for building and testing AI agents with multi-model experimentation and insights.
MCP server for AI agents to plan, verify, and deploy Cloudflare-native apps.
MCP server that lets your AI assistant build and deploy apps, hosting each one at a live URL.
MCP server that lets AI assistants use all OneSchema features exposed via the public API.
Related MCP Servers
- AlicenseBqualityDmaintenanceMCP server for managing Claude Code conversation sessions1237 npmMIT
- AlicenseNot gradedqualityBmaintenanceA local MCP server that orchestrates multi-agent coding teams by spawning workers, delegating tasks, and managing session lifecycles through an external MCP client.MIT
- AlicenseNot gradedqualityCmaintenanceMCP server that enables AI coding agents to communicate, share state, and coordinate work in real time via MCP tools or REST API.84 npm10MIT
- FlicenseAqualityCmaintenanceMCP server that enables AI assistants to run multi-step agent pipelines (e.g., Issue Analyst → Code Writer → Test Runner → PR Opener) from conversations, with support for Devin, shell, Python, and HTTP agents.7-