ashfulcra/fulcra-tools

Ash's Fulcra Tools

Vibe-coded by Fulcra's lawyer on Fulcra's own primitives — unofficial, unsupported, and a genuinely useful thing to point your agents at.

Note from the human: this is how I use Fulcra

I use Fulcra to build stuff and get stuff done by coordinating long-running agents across multiple platforms — Claude Code (desktop and cloud), Claude Cowork, OpenClaw, ChatGPT, and Codex. That includes having long-running coding-agent sessions capture my ideas, plan together, assign work to each other, and review each other's changes across model vendors (my review loop requires author and approver to be different models).

The agents coordinate all of that over Fulcra — a user-owned context backend that my agents and I use as a shared bus of typed events and documents — capturing what they're doing and session checkpoints along the way. Which means I can do things like review everything all the bots did over the past few days from my Fulcra dashboard, ask any of the bots what the others have done, or park a Claude Code cloud session midstream and pick it up in the Codex app.

Storing their state and work to Fulcra on every wake is what turned Claude Code cloud sessions into very long-running agents that behave a lot like OpenClaw or Hermes agents. And none of this needed anything but an ordinary Fulcra account: the bus is just custom data types and files in my own account — no coordination server, no broker, just reads and writes and some scaffolding the agents run in their sessions.

I'm letting the bots run wild, so I max out multiple foundation-model subscriptions' weekly limits — and the agents use Fulcra to load-balance across accounts (ATC, figuring this out). When something needs me, the fleet's watchdog sends a notification to my phone.

Those bots also built me a [