pipehero

Pipehero — MCP server + skill

Pipehero tunnels your localhost to a public URL and captures every webhook, so you can inspect, replay, and debug them — including from your AI agent via MCP.

This repo is auto-published from Pipehero's source. It holds the public MCP manifest and the agent skill. Issues/PRs welcome.

• 🌐 Website: https://pipehero.app
• 📚 Docs: https://pipehero.app/docs

Connect the MCP server

Let an agent (Claude Code, Cursor, Copilot, Windsurf, Zed, …) list your tunnels, read captured webhooks, and replay them to localhost.

Remote — recommended (no install, OAuth)

claude mcp add --transport http pipehero https://mcp.pipehero.app/mcp

Your client opens the browser to sign in and approve — no token to paste.

For CI/headless, pass a token from https://pipehero.app/dashboard/settings:

claude mcp add --transport http pipehero https://mcp.pipehero.app/mcp \
  --header "Authorization: Bearer <token>"

Local — via the CLI

curl -fsSL https://pipehero.app/install | sh
pipehero login
claude mcp add pipehero -- pipehero mcp

Any MCP client works — add it to .cursor/mcp.json, claude_desktop_config.json, etc.:

{ "mcpServers": { "pipehero": { "command": "pipehero", "args": ["mcp"] } } }

Tools

Tool	What it does
list_tunnels	Your tunnels and whether each is online.
list_requests(subdomain)	Recent captured webhooks for a tunnel.
get_request(subdomain, id)	Full request + response (headers + body).
replay_request(subdomain, id)	Replay a webhook to your localhost.
start_tunnel(name, port)	Expose a local port on a public URL (local MCP only).
stop_tunnel(name)	Stop a tunnel started with start_tunnel (local MCP only).
list_api_collections, get_api_collection(collection_id)	Your saved API docs: folders, endpoints, environments (remote MCP, and local from 0.1.14).
get_api_endpoint(collection_id, method + path)	One endpoint in full, with its version and response examples.
upsert_api_endpoint(collection_id, method, path, …)	Create or update an endpoint by route, so docs follow your code.
add_api_example(collection_id, method + path, name, status, …)	Save a response example.
delete_api_endpoint(collection_id, method + path)	Remove an endpoint whose route is gone.
import_api_spec(format, content) / export_api_collection(collection_id)	OpenAPI, Postman and curl, both ways.
run_api_endpoint(collection_id, method + path, environment)	Send it and read the real response.
save_captured_request(collection_id, tunnel, request_id)	Turn a real captured request into docs, without credentials.
check_api_drift(collection_id, tunnel)	Compare the docs with real traffic (Pro/Team).
list_api_monitors(collection_id)	Which endpoints are monitored and which are failing, with the last error (Pro/Team).
list_api_proposals	What happened to your proposed changes, on workspaces that review AI changes.
get_api_share_link(collection_id, method + path)	The endpoint's shareable link: a read-only page with its docs, examples and latest run.
share_api_endpoint(collection_id, method + path, …)	Create that link (workspace-only by default), open it to anyone with the link, or invite people by email.
share_api_collection(collection_id, …)	The same for a whole collection: every endpoint, read-only, never its environments.

Because your agent has both the captured webhook and your codebase, it can explain why a handler failed — then fix it and replay to confirm.

Skill

Install the skill so your agent knows when and how to use Pipehero:

pipehero skill install

or copy SKILL.md into ~/.claude/skills/pipehero/SKILL.md.

License

MIT