Build what you meant to , with agents that stay on track

The rinkata Hub UI is where your team and your AI agents share one source of truth: goals, specs, decisions, and completion evidence.

It works with the agents, GitHub, and review habits you already have. A continuous feedback loop keeps humans and agents aligned as the work evolves.

Get started
curl -fsSL https://rinkata.dev/install.sh | bash
Works withCodexClaudeChatGPTCursorGrokGemini
rinkata on Product Hunt

Watch the 50-second walkthrough

Intent thread

Idea Goal Work Decision Evidence

It gives your project a shared memory

All of your projects and even your company's documents live in one easy-to-access library.

  • A smart and searchable database to quickly find answers without sifting through pages of documents.
  • Completed work keeps its evidence attached to the goal.
  • Extract ideas from documents with one click.
  • Organize your documentation any way you please.

Knowledge · upload

Upload designs and docs about what to build

Bring in Figma exports, product briefs, and architecture notes. That becomes the shared project memory agents can read.

Input Designs, PRDs, and docs you upload
Output Searchable project memory agents can cite
  • AreaProduct
  • ExampleHandoff rules · UI principles · API notes
  • NextExtract or capture an Idea

It helps you capture your ideas efficiently

It has built-in tools to help you craft your ideas into workable goals.

  • A notebook drawer that's always available to jot down notes.
  • The ability to promote notes to ideas or add them to the knowledge base.
  • An agentic system to iron out your ideas through debate.
  • Auto-fill details from one line.
  • Add acceptance criteria from the beginning.

Ideas · capture → Goal

Capture an idea, promote it to a Goal

Lift a line from Knowledge, or jot a single idea. Shape it into a Goal with acceptance criteria before any agent starts.

Input A Knowledge excerpt, or one fresh idea
Output A Goal agents can claim and plan from
  • FromHandoff rules in Knowledge
  • GoalKeep handoff context across agents
  • NextPlan from Spec · claim in Work

It keeps your agents' work aligned with your intent

Connect your agent and get work done fast, with accuracy. Claim goals and send your agents to work. If a decision needs human intervention, you are empowered to make that call.

  • One board that shows progress for all work.
  • Quickly know if any decisions need humans, and make those decisions.
  • Supports work hand-offs between your team and its AI agents.
  • Take ownership of tasks.
  • See real-time progress of work being done.

Work · agents + judgment

Agents build; Needs attention holds judgment

Agents claim the Goal with Knowledge attached. When judgment is required, Needs attention surfaces the Decision. Humans Accept or Reject.

Input Goal + covering Knowledge
Output Progress, handoffs, and Needs attention Decisions
  • DecisionKeep path A or fold into path B?
  • AffectsHandoff context Goal
  • ActionHuman Accept / Reject

It produces proof when your work is complete.

Nothing is truly done until it's proven. It gathers the evidence for every completed goal, so you always know what landed and what evidence is attached.

  • Agents submit tests, screenshots, and pull requests as proof.
  • Add an optional human sign-off when you want a second set of eyes, never a blocker to deploy.
  • See what's accepted and what's still waiting on review at a glance.
  • Every result traces back to the goal and the decision behind it.

Example · Complete · evidence

Proof packs, attached to the goal

See what landed and the evidence attached to the goal.

Input Finished tickets under the Goal
Output Evidence pack
  • Tests24 passed · 0 failed
  • ScreenshotWork · handoff with goal attached
  • Pull request#1482 merged

Questions

  • Can I use it with my existing workflow?

    Yes. Keep your agents, GitHub, and review habits. Add rinkata as the shared memory and alignment layer so intent survives handoffs.

  • What is rinkata?

    It's a Hub for people and their AI agents. Knowledge, Ideas, Work, and Complete share one source of truth (Goals, specs, tickets, decisions, and evidence), so what you meant, what you committed to build, and what you deployed stay connected. It is not a replacement for your repo or your coding agent; it is the alignment layer that keeps everyone on the same outcome.

  • Does it write the code for me?

    No, you wire up your coding agents to do that.

  • How does it connect to coding agents?

    Through the rinkata CLI and agent integrations. Codex, Claude, Cursor, Grok, and Gemini can read and write project truth in Rinkata Hub UI so agents stay aligned with the same Goals, specs, and tickets.

  • Does it modify my repository?

    Hub owns Goals, specs, tickets, decisions, and evidence. Local files are cache or import/export surfaces, not a parallel source of truth agents should edit by hand.

  • Where is project knowledge stored?

    In rinkata UI: searchable Knowledge areas and documents, plus graph artifacts (Goals, specs, tickets). Agents can read covering docs over MCP and the CLI.

  • When does an agent need human approval?

    When judgment is required, especially Decisions on Needs attention. Agents may propose; humans Accept or Reject in rinkata UI.

  • What counts as completion evidence?

    Structured proof: test reports, screenshots, pull requests, and links attached at Complete. Optional human attestation is additive, never a silent gate that makes finished work look unfinished.

Built for the jobs around the agents

Same rinkata Hub UI. Different outcomes depending on who you are.

Preserve intent while agents build in parallel

Your north-star stays attached to every ticket and decision, even when five agents work at once.

Get started

Turn rough ideas into testable goals

Promote a notebook spark into a Goal with acceptance criteria before any agent starts.

Get started

See blockers and judgment calls

Needs attention surfaces Decisions that only a human can Accept, not silent drift between intent and work.

Get started

Hand work between agents without losing context

Handoffs keep the next agent on the same outcome, with evidence attached.

Get started