Rinkata
1 comment
Rinkata is a hub where teams and AI agents share one source of truth: goals, specs, decisions, and completion evidence. Knowledge uploaded to the hub — Figma exports, product briefs, architecture notes — becomes searchable memory that agents retrieve before inventing context. Capture an idea in the notebook drawer, promote it to a Goal with acceptance criteria, and agents claim it with the covering Knowledge attached. When judgment is required, Needs Attention surfaces the Decision and a human Accepts or Rejects. Finished work produces structured proof — test reports, screenshots, pull requests — and writes itself back into Knowledge so docs stay current. Handoffs keep the next agent on the same outcome with evidence attached. The CLI and agent integrations work with Claude, Codex, ChatGPT, Cursor, Grok, and Gemini over MCP. It is not a code writer, it is the alignment layer that keeps intent connected from idea to deploy. The Rinkata team used Rinkata to build Rinkata: 2,000+ tickets, 400+ specs, and 200+ decisions so far. Free for one person. Built by Bryan Schwab and the Quirence team.
Meet the team
What people who use it say
Nobody has reviewed Rinkata yet. If you have used it, you are the one who can say whether it holds up.
Write the first review
Sign in to comment
The problem most teams hit when they start running agents in parallel is that the agents invent context instead of reading what the team decided. Rinkata is the shared memory layer: goals, specs, and decisions live in one hub that agents search before they write code, so five agents working at once stay on the same outcome.
The part that made me stop was the Needs Attention queue. When an agent hits a judgment call — keep path A or fold into path B — it surfaces the Decision and waits for a human to Accept or Reject instead of picking one and moving on. That is the difference between agents that drift and agents that stay aligned. I have not run a real multi-agent project through it yet. The first thing I would check is whether the Knowledge retrieval actually prevents context invention, or whether agents still hallucinate details when the covering docs are incomplete and the task is ambiguous.