@happyrobot-ai/mcp

A Model Context Protocol (MCP) server for managing HappyRobot workflows. Works with Claude Desktop, Claude Code, or any MCP-compatible client.

Quick Start

Claude Code

claude mcp add -s user happyrobot -- npx @happyrobot-ai/mcp

Or with the API key pre-configured:

claude mcp add -s user happyrobot -e HAPPYROBOT_API_KEY=sk_live_xxx -- npx @happyrobot-ai/mcp

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "happyrobot": {
      "command": "npx",
      "args": ["@happyrobot-ai/mcp"],
      "env": {
        "HAPPYROBOT_API_KEY": "sk_live_your_key_here"
      }
    }
  }
}

.mcp.json (project-level)

{
  "mcpServers": {
    "happyrobot": {
      "command": "npx",
      "args": ["@happyrobot-ai/mcp"],
      "env": {
        "HAPPYROBOT_API_KEY": "sk_live_your_key_here"
      }
    }
  }
}

Configuration

| Variable | Description | Default | |---|---|---| | HAPPYROBOT_API_KEY | Your API key (sk_live_* or sk_test_*) | (none, required) | | HAPPYROBOT_API_BASE_URL | Override the API base URL | https://platform.happyrobot.ai/api/v2 |

If the API key is not set via environment variable, use the setup tool at runtime to provide it.

Available Tools

| Tool | Description | |---|---| | setup | Configure your API key at runtime | | create_workflow | Create a workflow from scratch, template, or inline nodes | | fix_broken_vars | Diagnose and fix broken variable references | | list_workflows | Search and list workflows with pagination | | get_workflow_details | Inspect workflow config, versions, and nodes | | update_workflow_nodes | Add, update, or delete nodes in a version | | manage_versions | Fork, publish, unpublish, lock, unlock versions | | trigger_run | Trigger a workflow run with payload | | monitor_runs | List runs, view sessions, recordings, manage annotations | | manage_variables | CRUD workflow-scoped environment variables | | manage_phone_numbers | List, purchase, update, and delete phone numbers | | list_integrations | Search integrations, events, and fetch integration resources (Slack channels, Google Sheets, etc.) | | get_available_variables | Get upstream variables available to a node | | get_node_config_schema | Get the configuration schema for a node's event | | manage_workflow | Cancel runs, update metadata/settings, delete, duplicate | | manage_sip_trunks | List, create, update, and delete SIP trunks | | browse_workflows | Browse workflow folders and explore folder hierarchy | | test_workflow | Test workflow versions — run all node tests or test a single node | | manage_mcp_servers | List, create, and refresh MCP server connections | | manage_credentials | List and create credentials for integrations | | fix_prompt_issues | Fetch prompt quality issues and get LLM guidance on fixing them | | manage_northstars | Manage northstar success criteria for evaluating prompt nodes | | manage_custom_evals | Manage custom eval tests (expected responses, tool calls) for prompt nodes | | manage_adversarial_tests | Manage adversarial (red-team) tests for agent nodes | | manage_adversarial_suites | Manage batched adversarial test suites for agent nodes | | get_node_details | Fetch full configuration details of a single workflow node |

Knowledge Base Prompts

The server also ships five wiki-style prompts that inject domain knowledge into the LLM context. Invoke them in Claude Code with /mcp__happyrobot__<name>:

| Prompt | Invocation | What it covers | |---|---|---| | Variable reference guide | /mcp__happyrobot__variable-guide | Syntax rules ({{index.field}} vs {{persistent_id.field}}), output path patterns per node type, common mistakes (AI Extract response. prefix, loop iteration_element), condition value format, diagnosing broken refs | | Node types reference | /mcp__happyrobot__node-types | What each node type does, key config fields, outputs, and common pitfalls | | Workflow tool patterns | /mcp__happyrobot__workflow-patterns | Tool sequencing, when to use get_node_details vs get_workflow_details, how fix_broken_vars and fix_prompt_issues differ, version management patterns | | Evaluation tools guide | /mcp__happyrobot__eval-guide | How northstars, custom evals, adversarial tests, and suites layer; when to use each; extracting evals from production runs | | Platform setup reference | /mcp__happyrobot__platform-setup | Integration/credential discovery, phone numbers, SIP trunks, MCP server connections, workflow variables |

API Documentation (Optional)

For full API docs access, add the HappyRobot docs MCP alongside this server. This lets Claude search and read the API reference when building workflows:

claude mcp add happyrobot-docs -- npx @mintlify/mcp --docs https://docs.happyrobot.ai

Development

# Install dependencies
npm install

# Run in development mode
HAPPYROBOT_API_KEY=sk_test_xxx npm run dev

# Build for production
npm run build

License

MIT