Connect an agent

1. Add the server

   Run this in your terminal:

   claude mcp add --transport http bountyhunt https://api.bountyhunt.xyz/mcp

   You’ll see “Added HTTP MCP server bountyhunt…” — that just saves the server to Claude Code’s config. Nothing connects yet; that happens inside Claude Code in the next step.

2. Start Claude Code and sign in

   From the same directory where you ran the command above (the server is saved per-project), start Claude Code with claude, then type /mcp, pick bountyhunt, and choose Authenticate. A browser opens to sign in with Google or GitHub — that authorizes the connection, with no key to copy or store.

   (Skipping /mcp also works: the first time the agent calls a BountyHunt tool, the same sign-in opens.)

3. Ask your agent

   That’s it. Try it with:

   What’s new across my BountyHunt programs?

Using another agent?

Cursor, Windsurf, or any MCP-capable client — paste this and let it wire up the server itself, then sign in when the browser opens.

Add the BountyHunt MCP server (streamable HTTP) at https://api.bountyhunt.xyz/mcp. It uses OAuth — a browser window will open for me to sign in. Once connected, call list_activity and show me the latest events.

Troubleshooting

If the agent can’t connect, authentication is stuck, or tool calls keep failing, the cleanest fix is to remove the server and add it back fresh. Run this from the same directory where you first added it (the server is saved per-project):

1. Remove the server

   claude mcp remove bountyhunt

2. Confirm it's gone

   claude mcp list

   bountyhunt should no longer appear in the list.

3. Add it again

   claude mcp add --transport http bountyhunt https://api.bountyhunt.xyz/mcp

   Then restart Claude Code, run /mcp, pick bountyhunt, and choose Authenticate to sign in again.

Reference

MCP access is part of BountyHunt Pro. Free accounts can connect, but tool calls return an upgrade prompt instead of data — upgrade to enable them.

Endpoint — https://api.bountyhunt.xyz/mcp (streamable HTTP).

Tools — list_activity, activity_digest, list_programs, get_program, list_assets, get_asset, list_audits, list_webhooks, watch_program, unwatch_program, request_feature, report_bug.

The agent acts as you: it sees exactly the data you do, and writes (feature requests, bug reports) are attributed to your account.