@junct/cli

Make your stack agent-ready — generate MCP servers and multi-format agent tools from any API spec, smart contract ABI, or llms.txt.

Install

npm install -g @junct/cli
# or
npx @junct/cli <command>

Quickstart

# Generate an MCP server + all agent-framework outputs from an OpenAPI spec
junct generate https://hermes.pyth.network/docs/openapi.json --id pyth

# Submit to registries (MCP Registry, Smithery, etc.)
junct submit pyth

# Inspect a server
junct status pyth

# List your fleet
junct list

# Deep-validate every tool
junct validate pyth

Commands

| Command | Purpose | |---|---| | junct generate <url> --id <clientId> | Generate MCP + multi-format tools from a spec URL | | junct submit <clientId> [--venues v1,v2] | Submit to one or more registry venues | | junct status <clientId> | Show server metadata, tools, and submission state | | junct list | List every server you have generated | | junct validate <clientId> | Deep-validate — handshake + execute every tool | | junct formats | Show supported output formats |

Output formats

mcp, openai, gpt-actions, langchain, eliza, agentkit, vercel-ai, crewai. Pass --outputs mcp,openai to limit.

Flags

| Flag | Purpose | |---|---| | --id | Client identifier (lowercase, alphanumeric, hyphens) | | --name | Display name (defaults to clientId) | | --outputs | Comma-separated output formats (default: all) | | --api-key | API key injected into generated server code | | --venues | Submission venues for submit | | --api | Override API base URL (default https://api.junct.dev) | | -h, --help | Help | | -v, --version | Version |

Hosted backend

The CLI is a thin client over https://api.junct.dev. Every generated MCP gets:

• Live endpoint at https://{clientId}.mcp.junct.dev/mcp
• Dashboard page at https://junct.dev/servers/{clientId}
• 8 output format downloads (MCP, OpenAI, LangChain, Eliza, AgentKit, CrewAI, GPT Actions, Vercel AI)

License

MIT