Enoch

enoch-mcpis a local Model Context Protocol stdio server for a configured Enoch FastAPI control-plane API.It lets MCP clients inspect and operate Enoch through typed tools without exposing a raw shell or reimplementing Enoch business logic.

- Registers MCP tools for Enoch control-plane, Dashboard V1, and core endpoints.
- Sends requests to a configured Enoch API URL.
- AddsAuthorization: Bearer <token>to API requests using the configured token.
- Returns Enoch API responses to the MCP client.
- Marks read-only tools with MCP read-only annotations.
- Marks mutating tools as non-read-only and adds approval metadata.
- Keeps safe defaults for dry-run operations.
- Optionally probes configured CPU/GPU workers directly through worker APIs or allowlisted SSH diagnostics.

- It does not expose a raw shell tool.
- It does not accept arbitrary SSH commands.
- It does not read or write local artifact files directly.
- It does not call language models.
- It does not cache, retry, queue, or schedule work.
- It does not bypass Enoch authentication or authorization.

- Python 3.11 or newer
- A running Enoch API, normally athttp://localhost:8787
- An Enoch API bearer token
- An MCP client that can run local stdio servers

uvx enoch-mcp --api-url http://localhost:8787 --api-token '<token>'

Or configure with environment variables:

export ENOCH_API_URL='http://localhost:8787' export ENOCH_API_TOKEN='<token>' uvx enoch-mcp

git clone https://github.com/alias8818/enoch-mcp.git cd enoch-mcp uv sync --dev uv run enoch-mcp --api-url http://localhost:8787 --api-token '<token>'

The token is required. If it is missing, tool calls fail before making an HTTP request.

Worker probes are disabled unlessENOCH_WORKER_PROBES_JSONorENOCH_WORKER_PROBES_FILEis configured. This keeps the default package a thin control-plane bridge. When configured, the MCP exposes named diagnostics for worker truth: API health, worker-gate dashboard status, active process markers, bounded log tails, disk space, and expected artifact presence.

{ "cpu": { "api_url": "http://127.0.0.1:18788", "api_token": "worker-api-token", "service_name": "enoch-control-plane", "project_root": "/srv/enoch/projects" }, "gpu": { "api_url": "http://127.0.0.1:18789", "api_token": "worker-api-token", "ssh_host": "worker-gpu.example.internal", "ssh_user": "enoch", "service_name": "enoch-control-plane", "project_root": "/srv/enoch/projects", "log_paths": ["/var/log/enoch-control-plane.log"] } }

- api_url: worker-gate base URL. Used first for/healthz,/dashboard/api,/dashboard/api/run/{run_id}, and/project-status/{project_id}.
- api_token: worker bearer token. Treated as secret.
- ssh_host,ssh_user,ssh_port: optional SSH fallback/debug target.
- service_name: systemd unit name for service checks and journal tails.
- project_root,state_dir: fixed worker roots used for disk and artifact checks.
- log_paths: fixed worker-gate log paths that may be tailed.

SSH probes only run fixed diagnostic commands. They do not accept arbitrary shell input from the MCP client. User-supplied IDs are limited to safe run/project identifier characters, log output is bounded, SSH uses batch mode and no stdin, and the recommended deployment is a read-only worker user or forced-command policy.

{ "mcpServers": { "enoch": { "command": "uvx", "args": ["enoch-mcp"], "env": { "ENOCH_API_URL": "http://localhost:8787", "ENOCH_API_TOKEN": "replace-with-token" } } } }

For local development, point the MCP client at the checkout:

{ "mcpServers": { "enoch": { "command": "uv", "args": ["--directory", "/path/to/enoch-mcp", "run", "enoch-mcp"], "env": { "ENOCH_API_URL": "http://localhost:8787", "ENOCH_API_TOKEN": "replace-with-token" } } } }

Use the equivalent local stdio-server settings for Cursor, Copilot, Windsurf, or other MCP clients.

uv sync --dev uv run pytest -q gitleaks detect --no-git --redact

This package is public. Examples must use placeholders and local URLs only. Do not publish private hostnames, LAN/Tailscale IPs, operator paths, or live tokens.

This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.

AI-powered Jenkins pipeline intelligence platform with natural language interface. Provides comprehensive pipeline analysis, failure prediction, optimization suggestions, and automated Jenkinsfile reconstruction using Model Context Protocol (MCP) integration.

Stateful health monitoring, diagnostics, and web attestation for AI agents. 11 MCP tools. Free Founder's Beta

Paid remote MCP for CLI tool MCP, structured receipts, usage logs, and audit-ready evidence for agent and CI workflows.

Remote MCP server that gives your AI agent a dead-man's-switch: it calls create_heartbeat, pings on schedule, and emails you the instant an agent or cron loop stops checking in.

AI agent infrastructure suite — task deduplication, cost prediction, output validation, behavioral governance, shared failure intelligence. 5 products, 14 MCP tools, 1 API key.

Manage and observe Prefect workflows through natural language.

A modern, extensible multi-agent workflow engine with real-time monitoring and a web visualization interface.

MCP server for catching cron silent failures — jobs that exit 0 with empty output, retry storms, action-budget leaks. 6 silent-fail patterns across system cron, systemd timers, OpenClaw cron logs.

Cron scheduling, uptime monitoring, heartbeat monitoring, and email verification as native MCP tools for Claude Code, Cursor, and any MCP-compatible agent.

GPU infrastructure control plane + per-agent LLM FinOps. 74 tools. Deploy, scale, track cost per agent, enforce budgets. MIT license.