Carbon Forge as an MCP tool for LlamaIndex agents: one versioned record for code and docs

[Theaxiom]

LlamaIndex agents read documents well; when they also write, the team needs the pages and code they change kept with a history of what changed and why. Carbon Forge keeps code, files and prose in one record on the server, and a code change and the page that explains it land in one changeset.

It is a remote MCP server at https://carbonforge.dev/api/mcp (Streamable HTTP), so the route is llama-index-tools-mcp with McpToolSpec. Public workspaces read without a token; writing needs a bearer token. I have not yet run it from a LlamaIndex agent.

Blame, grep, diff and history run on the server, and blame shows the intent the agent wrote beside each line, so a reviewer can ask why a line exists and get the agent's own sentence.

We would like to build a worked example with anyone here who wants one, and we offer three-month Carbon Forge subscriptions for that team and a batch for this community, with a promo code agreed with me.

Would a notebook in the docs or a community example be the better home for it?

Docs: https://carbonforge.dev/docs/p/home?utm_source=llamaindex&utm_medium=referral&utm_campaign=partners-2026-10-28

Disclosure: I build Carbon Forge at 100monkeys.ai, and this post was prepared by an AI agent working for me.