The official MCP server for Strait -- an open-source job execution and workflow orchestration platform for humans and AI agents. Lets any MCP-compatible client (Claude Desktop, Cursor, Windsurf, or custom agents) manage the complete Strait platform natively.

133 tools, 5 resources, 9 prompts. Full API coverage. Built with Effect TS for typed errors, structured concurrency, and dependency injection.

Website | Platform Repo | Documentation | npm

Installation

npm (recommended)

npm install -g @strait/mcp-server

npx (no install required)

npx @strait/mcp-server

Docker

docker pull ghcr.io/strait-dev/mcp

Requires Node.js 20+ for npm/npx.

Quick Start

Set your API key:

export STRAIT_API_KEY=strait_live_...

Run in stdio mode (for IDE integrations):

npx @strait/mcp-server

Run in HTTP mode (for remote/production use):

npx @strait/mcp-server --transport http --port 3000

Client Setup

Claude Desktop

Add to your Claude Desktop config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "strait": {
      "command": "npx",
      "args": ["@strait/mcp-server"],
      "env": {
        "STRAIT_API_KEY": "strait_live_..."
      }
    }
  }
}

Restart Claude Desktop after saving. The Strait tools will appear in the tools panel.

Claude Code

Add via CLI (recommended):

claude mcp add strait --scope user -- npx @strait/mcp-server

Or add to ~/.claude.json manually:

{
  "mcpServers": {
    "strait": {
      "command": "npx",
      "args": ["@strait/mcp-server"],
      "env": {
        "STRAIT_API_KEY": "strait_live_..."
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project root (project-level) or ~/.cursor/mcp.json (global):

{
  "mcpServers": {
    "strait": {
      "command": "npx",
      "args": ["@strait/mcp-server"],
      "env": {
        "STRAIT_API_KEY": "strait_live_..."
      }
    }
  }
}

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "strait": {
      "command": "npx",
      "args": ["@strait/mcp-server"],
      "env": {
        "STRAIT_API_KEY": "strait_live_..."
      }
    }
  }
}

OpenAI Codex CLI

Add via CLI:

codex mcp add strait -- npx @strait/mcp-server

Or add to ~/.codex/config.toml:

[mcp_servers.strait]
type = "stdio"
command = "npx"
args = ["@strait/mcp-server"]

[mcp_servers.strait.env]
STRAIT_API_KEY = "strait_live_..."

VS Code + GitHub Copilot

Open Settings (Ctrl+Shift+P > "MCP: Add Server") or add to .vscode/mcp.json:

{
  "servers": {
    "strait": {
      "command": "npx",
      "args": ["@strait/mcp-server"],
      "env": {
        "STRAIT_API_KEY": "strait_live_..."
      }
    }
  }
}

Zed

Add to Zed settings (Zed > Settings > Open Settings):

{
  "context_servers": {
    "strait": {
      "command": {
        "path": "npx",
        "args": ["@strait/mcp-server"],
        "env": {
          "STRAIT_API_KEY": "strait_live_..."
        }
      }
    }
  }
}

Continue.dev

Add to ~/.continue/config.yaml:

mcpServers:
  - name: strait
    command: npx
    args: ["@strait/mcp-server"]
    env:
      STRAIT_API_KEY: strait_live_...

Cline

Click the MCP Servers icon in Cline's toolbar, then "Edit MCP Settings" and add:

{
  "mcpServers": {
    "strait": {
      "command": "npx",
      "args": ["@strait/mcp-server"],
      "env": {
        "STRAIT_API_KEY": "strait_live_..."
      }
    }
  }
}

Amazon Q Developer

In VS Code or JetBrains, open Amazon Q settings and add an MCP server with:

• Command: npx
• Args: @strait/mcp-server
• Environment: STRAIT_API_KEY=strait_live_...

JetBrains IDEs

Go to Settings > Tools > AI Assistant > MCP Servers, click "+" and add:

• Command: npx
• Args: @strait/mcp-server
• Environment: STRAIT_API_KEY=strait_live_...

OpenCode

Add via CLI:

opencode mcp add strait -- npx @strait/mcp-server

Or add to your OpenCode config:

mcp:
  strait:
    type: local
    command: npx
    args: ["@strait/mcp-server"]
    env:
      STRAIT_API_KEY: strait_live_...

Docker

For remote or production deployments using HTTP transport:

docker run -e STRAIT_API_KEY=strait_live_... -p 3000:3000 ghcr.io/strait-dev/mcp

Or with Docker Compose:

services:
  strait-mcp:
    image: ghcr.io/strait-dev/mcp
    ports:
      - "3000:3000"
    environment:
      STRAIT_API_KEY: ${STRAIT_API_KEY}

What You Can Do

Once connected, ask your AI client to interact with Strait:

Jobs -- "List my jobs", "Create a job called data-sync that hits https://api.example.com/sync every hour", "Trigger the etl-pipeline job with this payload", "Show version history for my deploy job"

Runs -- "Show logs for the last failed run", "Why did run run_abc123 fail?", "Watch this run until it finishes", "Compare the last two runs of my sync job"

Workflows -- "Visualize the DAG for my nightly-batch workflow", "Create a workflow with three steps and an approval gate", "Trigger the onboarding pipeline"

Deployments -- "Validate this config before deploying", "Show what would change if I deploy", "Roll back the last deployment"

Diagnostics -- "What is the overall health of my project?", "Debug why run run_abc123 failed", "Suggest cost optimizations for my etl job"

Analytics -- "Show me cost trends for the last 30 days", "What are my top failing jobs?", "Show job reliability metrics"

Infrastructure -- "List my webhook subscriptions", "Create a notification channel for Slack", "Show my environments"

Security -- "List my API keys", "Rotate the deploy key with a 24-hour grace window", "List my secrets"

Configuration

| Option | Env Var | CLI Flag | Default | |--------|---------|----------|---------| | API Key | STRAIT_API_KEY | --api-key | (required*) | | Base URL | STRAIT_BASE_URL | --base-url | https://api.strait.dev | | Transport | - | --transport | stdio | | HTTP Port | - | --port | 3000 | | OAuth Issuer | OAUTH_ISSUER_URL | - | (disabled) | | OAuth Resource | OAUTH_RESOURCE_URL | - | (disabled) |

*API key is optional when OAuth is configured for HTTP transport. Set OAUTH_ISSUER_URL and OAUTH_RESOURCE_URL to enable OAuth -- users authenticate via browser instead of API key.

Features

Tools (133)

All operations return AI-optimized summaries with temporal context, resource IDs, and next-step hints.

| Category | Count | Examples | |----------|-------|---------| | Jobs | 15 | list, get, create, update, delete, trigger, bulk trigger, versions, clone, health, batch create/enable/disable, dependencies | | Runs | 19 | list, get, logs, events, cancel, retry, watch, last run, diff, DLQ, bulk replay, children, outputs, usage, tool calls, pause, resume, reschedule | | Workflows | 15 | list, get, create, update, delete, trigger, visualize DAG, graph, clone, dry-run, plan, simulate, versions, version diff | | Workflow Runs | 15 | list, get, cancel, steps, watch, approve, reject, graph, timeline, resume, retry, force-complete step, retry step, bulk cancel, explain | | Event Triggers | 3 | list, get, send event | | Secrets | 3 | list (names only), create, delete | | Deployments | 7 | deploy, list, rollback, validate, diff, finalize, promote | | API Keys | 4 | list (masked), create, revoke, rotate | | Job Groups | 6 | list, get, create, update, delete, stats | | Environments | 5 | list, get, create, update, delete | | Projects | 4 | list, get, create, settings | | Webhooks | 7 | list, get, create, delete, deliveries, retry delivery, test | | Analytics | 10 | performance, costs, cost trends, run summary, failure reasons, job history, reliability, top failing, workflow completion rates, step durations | | Usage | 5 | current usage, history, forecast, project costs, spending anomalies | | Event Sources | 5 | list, get, create, update, delete | | Notifications | 4 | list, get, create, delete | | Log Drains | 4 | list, get, create, delete | | Operations | 2 | project status dashboard, OpenAPI spec |

Resources (5)

| URI | Description | |-----|-------------| | strait://connection | Connection status, auth verification, API key scopes | | strait://health | Infrastructure health: job counts, failure rate, scheduled runs | | strait://jobs | Browseable list of all jobs with status, schedule, and run health | | strait://workflows | Browseable list of all workflows with step count and status | | strait://analytics | Current cost and performance snapshot (success rate, spend, top failures) |

Prompts (9)

| Prompt | Arguments | Description | |--------|-----------|-------------| | debug-failure | run_id | Root cause analysis: fetches run + logs, reports what failed, why, and suggested fix | | optimize-cost | job_slug | Cost analysis: reviews run history, suggests timeout/retry/concurrency optimizations | | incident-response | (none) | Production incident investigation: check health, find failures, analyze root causes, recommend fixes | | setup-job | description | Guided job creation from natural language: infers config, creates job, runs test, sets up monitoring | | cost-report | period? | Comprehensive cost report: spend trends, top cost drivers, failure waste, projected costs | | workflow-status | workflow_run_id | Workflow run deep-dive: step-by-step breakdown, bottlenecks, failure analysis | | deploy-preview | config | Guided deployment: validate, diff, deploy with confirmation, monitor results | | onboard-project | name, description? | New project setup: create project, test job, verify connectivity, suggest next steps | | security-audit | (none) | Security review: API key health, secrets, webhook HTTPS compliance, spending anomalies |

Error Handling

All errors are typed using Effect's tagged error pattern with automatic retry for transient failures:

| Error | Trigger | Response | |-------|---------|----------| | NetworkError | API unreachable | Base URL hint, check connectivity | | ApiError | Non-2xx response | Status-specific hints (404: use list tool, 401: check scopes) | | RateLimitError | 429 response | Parsed Retry-After header, automatic exponential backoff | | ValidationError | Invalid input | Field name, constraint, example | | PollTimeoutError | Watch timeout | Resource ID, elapsed time |

Response Shaping

Every tool response is AI-optimized:

• Resource IDs in copyable form (run_abc123)
• Next-step hints ("Use strait_get_run with ID 'run_abc123' to check progress")
• Temporal context ("2m ago") instead of raw timestamps
• Lists capped at 25 items with cursor-based pagination
• Null/empty fields omitted

Architecture

MCP Client <-- stdio or Streamable HTTP --> strait-mcp (Node.js) -- REST API --> Strait API

Built with:

• MCP SDK -- official TypeScript SDK for Model Context Protocol
• Effect -- typed errors, dependency injection, structured concurrency
• Zod -- tool input schema validation (MCP SDK boundary)

Development

bun install              # install dependencies
bun run dev              # stdio mode
bun run dev:http         # HTTP mode
bun run test             # run unit tests (304+ tests)
bun run test:integration # run integration tests (Docker required)
bun run typecheck        # type check with tsgo
bun run biome:lint       # lint with biome + ultracite
bun run biome:format     # format
bun run run-all          # lint + format + typecheck

Project Structure

src/
  index.ts               -- entrypoint, transport selection, Effect runtime
  server.ts              -- McpServer creation, tool/resource/prompt registration
  config.ts              -- Effect config service
  errors.ts              -- tagged error taxonomy (NetworkError, ApiError, ...)
  schema.ts              -- slug, cron, step type validators
  types.ts               -- Strait domain model types
  client/
    service.ts            -- StraitClient Effect service tag
    http.ts               -- fetch-based implementation with retry/backoff
    jobs.ts ... apikeys.ts -- domain API functions returning Effects
  tools/
    helpers.ts            -- runEffect bridge, response shaping, formatError
    jobs.ts ... operations.ts -- MCP tool handlers
  resources/              -- strait://connection, strait://health
  prompts/                -- debug-failure, optimize-cost
tests/                   -- 221 tests with mock HTTP server

Troubleshooting

"STRAIT_API_KEY is required"

Set the environment variable before running, or pass it via the env block in your IDE config. Keys start with strait_live_ or strait_test_.

"NetworkError: API unreachable"

Check that STRAIT_BASE_URL is correct (defaults to https://api.strait.dev). If running behind a proxy, ensure outbound HTTPS is allowed.

"ApiError: 401 Unauthorized"

Your API key may be expired or missing required scopes. Generate a new key at strait.dev with the scopes needed for the tools you are using.

"ApiError: 404 Not Found"

The resource (job, run, workflow) does not exist or you lack access. Use the corresponding list tool to see available resources.

"RateLimitError: 429"

The server retries automatically with exponential backoff. If this persists, reduce request frequency or contact support.

MCP client does not detect the server

1. Verify the config file path is correct for your OS
2. Ensure npx is on your PATH (which npx to check)
3. Restart your IDE after changing MCP config
4. Check IDE logs for connection errors

Docker container exits immediately

1. Ensure STRAIT_API_KEY is set: docker run -e STRAIT_API_KEY=... ghcr.io/strait-dev/mcp
2. Check logs: docker logs <container-id>
3. The default transport is HTTP on port 3000 -- ensure the port is not in use

License

MIT