Give your agent the keys to the project
Pibox runs a Model Context Protocol server. Point Claude, Cursor or VS Code at it and the agent works with your files, threads and tasks directly — no glue code.
MCP endpoint
https://mcp.pibox.com/mcpIt authenticates with one of your personal API tokens as a Bearer header. Create a token in Settings › API tokens — it is shown once, so copy it straight away.
Paste it into your client
Replace <YOUR_TOKEN> with your own token.
claude mcp add --transport http pibox https://mcp.pibox.com/mcp \
--header "Authorization: Bearer <YOUR_TOKEN>"Treat the token like a password. A personal token carries full access to your account — keep it out of shared repos and group chats, and delete any token you stop using from the same panel.
What the agent can reach
Everything the public API exposes — the agent is not working from a scraped page, it is calling the same endpoints the app does.
Find the file without being told where it is
Search across projects by filter instead of pasting paths. “The latest bounce of that track” resolves to an actual file.
Read the review thread
Pull chat history for a project, so the assistant answers with what the client actually said rather than a summary you retyped.
Turn notes into tasks
Create tasks on a folder, or load a saved template, so a round of feedback becomes a checklist without anyone transcribing it.
Keep the project tidy
Create folders, move things in, update properties. The boring half of every delivery, handled by something that does not get bored.
Tag audio automatically
Send a track and get AI tags back — the same endpoint the app uses, now available to your agent.
React to what changes
Webhooks fire on uploads, renames, status changes and closed tasks, so an agent can act the moment a mix is approved.
Building without an agent? The API docs cover every endpoint, and they publish an llms.txt index — a plain-text map an assistant can read on its own.
Questions
What is the Pibox MCP endpoint?
https://mcp.pibox.com/mcp — it speaks the Model Context Protocol over HTTP and authenticates with one of your personal API tokens as a Bearer header.
What is MCP?
The Model Context Protocol is an open standard for connecting AI assistants to external tools and data. A client that speaks it — Claude, Cursor, VS Code and a growing list of others — can use any MCP server without a custom integration.
Where do I get a token?
In the Pibox app open Settings › API tokens, under Developer, and click Create token. It is shown once, so copy it straight away.
What can the agent actually do?
Whatever the public API covers: teams and projects, file search and upload, folders, tasks and task templates, chat history and messages, AI track tagging, and webhook endpoints. The full reference is on the docs site.
How much access does a token carry?
Personal tokens carry full access to your account, so treat them like passwords: keep them out of shared repos and chat, and delete any token you no longer use from the same Settings panel.
Do my teammates need their own tokens?
Yes. A token acts on behalf of the person who created it, so each teammate connecting an agent creates their own from their own account.
Connect it and ask it something
Paste the config, then try “what changed in this project today?” — the answer comes from your actual files.