Labsco
MCP SERVER

Pickaxe AI Agent MCP

by aplaceforallmystuff

Read what users actually asked a Pickaxe agent, fix the knowledge-base gaps that surface, and manage the studio's users and product access from the same client.

Customer Support & HelpdeskVerified
Summary
It is built around one loop — read chat_history, find what the agent could not answer, then doc_create and doc_connect to close it.

chat_history's own description names that job: analyse user questions, identify knowledge-base gaps, review agent performance. The document calls are shaped to answer it, with doc_create accepting either raw content or a website URL to scrape. The user side is full write access rather than reporting — user_update sets products, currentUses and extraUses — and one consequence is worth knowing before a cleanup pass: doc_delete removes the document from every agent connected to it, not only from the one you were looking at.

What it is

A management surface for a Pickaxe studio: an agent's conversation history, the documents in its knowledge base, the studio's users and their product access, and the memories collected against them.

What you get
  • Conversation history you can act on: chat_history fetches an agent's conversations by pickaxeId with limit, skip and a format choice, for reviewing questions asked and performance given.
  • Knowledge-base documents created either way: doc_create takes rawContent or a website URL to scrape, doc_list pages the studio's documents, doc_get returns one by documentId, and doc_delete removes it from every agent it was connected to.
  • Per-agent knowledge wiring: doc_connect links a document into a Pickaxe agent's knowledge base and doc_disconnect unlinks it, so one document can serve one agent and not another.
  • User administration inside the studio: user_list returns users with their product access and usage stats, user_create and user_update set name, password, isEmailVerified and products, user_invite emails invitations with optional productIds, and user_get and user_delete work by email address.
  • Usage counters as editable fields: user_update takes currentUses and extraUses alongside the user's product list.
  • What the agent has remembered: memory_list returns the studio's memory schemas and memory_get_user returns one user's collected memories for a memoryId.
  • More than one studio behind one server: studios_list reports the configured Pickaxe studios and which is the default, and a studio argument on a call sends it somewhere else.
  • The product catalogue behind access grants: products_list returns the studio's available products and bundles, which are what user_create, user_update and user_invite assign.
Requirements

A Pickaxe studio to act against, and the pickaxeId of the agent whose history and knowledge base you want to reach.

Setup effort

One command — npx -y mcp-pickaxe