Rinkata Review (2026): Pricing, Features & Honest Verdict

TLDR

Rinkata is a hosted hub where a software team and its coding agents share one record of plans, work, decisions and proof of completion. The idea is clever and the policies are unusually honest, but it is young and opinionated. Best for: small teams running several agents on one product. Price: from $15/mo per seat, free plan yes. Rating: 6.5/10.

What is Rinkata?

We found Rinkata on Product Hunt, where it launched on 30 September 2026. The pitch is one source of truth for a team and its agents. In practice, that means a web app called Rinkata Hub where people write down what they want built, agents pick up the work, and finished work comes back with proof attached.

It is not a coding assistant and it does not touch your repo directly. The FAQ says so plainly: you wire up your own agents to write code. Rinkata sits beside Claude, ChatGPT, Grok, Codex and a popular AI code editor, holding the tickets, specs and decisions those agents read before they start. The service is run by Quirence, Inc., a Menlo Park, California company, and is hosted only, with no self-hosted edition.

What Are Rinkata’s Key Features?

Four views that follow a piece of work from idea to proof

The app is organised as Knowledge, Ideas, Work and Complete. Knowledge holds uploaded documents such as PRDs, Figma exports and architecture notes. Ideas is a scratchpad where a one-line thought can be promoted into a goal with acceptance criteria. Work shows tickets in flight, grouped into Needs attention, In progress and Ready to start. Complete gathers the proof. Each item keeps a link back to the goal and decision behind it, which is the whole point of the product.

A large MCP tool set with human-only stamps

Agents connect over MCP and get a long list of rinkata_ tools, from rinkata_search and rinkata_work_queue to rinkata_claim and rinkata_complete_ticket. The docs mark each tool as You, Agent or Either. Approving a spec, ratifying a decision and applying a reconcile plan are human-only, so an agent can propose but cannot sign off its own work. The docs also direct agents to search before creating anything, which targets the duplicate-ticket mess that multi-agent setups tend to produce.

Completion proof that is checked, not only attached

An agent finishes a ticket by attaching a PR, a commit, a real test run or screenshots. The What’s new page shows the team tightening this: since 24 September 2026 the app refuses screenshots smaller than 1440 by 900 or visibly soft. When a human accepts a spec, Rinkata can generate manual test cases and write a summary back into Knowledge, so documentation grows from finished work rather than from someone remembering to update it.

Pressure debates for rough ideas

Pressure is an optional feature that has two AI models argue opposite sides of an idea before you commit to it. The privacy policy says the topic and context are truncated to 500 characters and sent to xAI (Grok) and Google Gemini. It is a clever way to stress an idea, though it is also the only feature that sends your content to xAI.

What Is It Like to Use Rinkata?

Getting started happens in a browser: sign in with Google or an email link, create a project, then connect an agent. For Claude Desktop or claude.ai you add a custom connector pointing at api.rinkata.dev/mcp and approve a sign-in, with no key involved. Claude Code uses a terminal sign-in. ChatGPT connects through OAuth and each teammate authorises separately. Grok and the code editor use a minted provider key.

There is also a command-line tool installed with a curl script. We read the script: it verifies an Ed25519 signature on the release list before unpacking, refuses to run as root, and ships builds for macOS and Linux only, on both ARM and x64. The signed release list pointed at version 0.8.16 when we checked. Running rinkata install afterwards detects Claude Code, Codex, Antigravity, OpenClaw and Grok setups, among others.

The vocabulary is the steepest part. The docs talk about graph hashes, staged reconcile plans, successor specs and ratified decisions, and the changelog references internal IDs like DEC-PLUMB-220. The installer still drops legacy plumb command names, a leftover from an earlier product name. None of that is fatal, but a team that only wants a to-do list for its agents will find the first hour dense.

Where Does Rinkata Fall Short?

Your content goes to Google Gemini for chat, ticket drafting, status summaries and knowledge builds. The privacy policy says a knowledge build reads every markdown file in every attached repository plus every artifact in the project, calling it the largest transfer in the product. It also states Rinkata uses a paid Gemini tier that Google does not train on. That is a clear disclosure, but some teams will not want their repos read by a third-party model at all.

Data portability is thin. The privacy policy says there is no self-service export and no self-service account deletion; every request goes to a person by email with a 30-day response window. Data lives in a single Google Cloud region, us-central1, with no EU residency option. For a tool that wants to be your project memory, the lack of a one-step export is the biggest gap.

The marketing site overpromises a little. It shows Codex and Gemini logos, yet the docs only carry setup guides for Claude, ChatGPT, Grok and the code editor; Codex and Gemini’s Antigravity arrive through the CLI instead. The site also claims Rinkata removes the need for tech writers, which is a big claim for a product at version 0.8.

Billing has rough edges too. Seats follow workspace membership automatically, and the terms say Rinkata does not ask you to confirm the change first. Plan and price changes are not self-service; you email support. Rinkata also collects no sales tax or VAT, so buyers handle any tax that applies themselves.

How Much Does Rinkata Cost?

Rinkata starts at $0 on a free plan and $15 per seat per month for Team. Annual Team billing is $150 per seat per year, about $12.50 a month. There is no free trial of Team; the free plan is the trial, and the pricing page says it gives access to all of Rinkata.

The free plan allows one seat, 250 tickets, 100 specs, 100 decisions and 1,500 artifacts in total, plus 25 AI calls and 2 knowledge builds a month, each build reading up to 1,000 sources. Team lifts item limits to unlimited within a 50,000 artifact total, with 500 AI calls, 30 builds of up to 5,000 sources, and unlimited seats. An Enterprise tier is quoted by the sales team.

Two details only appear in the Plans and limits doc. Debates are metered separately, 3 a month on Free and 30 on Team. And item limits count everything that exists, so archiving a ticket does not hand the allowance back. The doc defines an AI call as one request to Ask, drafted tickets or contextual help. Your own agents reading and writing over MCP are not listed as AI calls.

Against general trackers, Rinkata is pricier. Linear Basic is $10 per user per month billed yearly, Plane Pro is $6 per seat per month billed annually, and Jira Standard is $7.91 per user per month. Rinkata charges for the agent-alignment features those tools do not center on.

Refunds are fair. Monthly plans get no partial refunds, while annual plans are refunded in full if you cancel within 14 days of the first payment. Duplicate charges or charges after cancelling are always refunded, and dropping to Free deletes nothing.

Who is Rinkata Best For?

Use Rinkata if you run two or more agents on the same product and keep losing the thread between sessions. It suits founders and small engineering teams who already write specs, want an agent to check the plan before it starts, and want a human sign-off on the calls that count. The free plan is generous enough to test that on a real side project.

Skip Rinkata if your team needs EU data residency, self-service export or a self-hosted option. Skip it too if your agents live on Windows machines, since the CLI ships only for macOS and Linux, or if you want a general project tracker for non-engineers. Linear or Jira will do that job for less.

Best Rinkata Alternatives

Linear

Linear is the polished issue tracker many product teams already use, and it now lists an agent platform on its pricing page. The Free plan covers 2 teams and 250 issues, Basic is $10 per user per month billed yearly, and Business is $16. Pick it if you want a mature tracker first and agent features second.

Plane

Plane is an open-core tracker with cloud and self-hosted editions. Free covers up to 12 users, Pro is $6 per seat per month billed annually with 500 AI credits per seat, and Business is $13 on the same annual terms. It is the pick for teams that need to host their own data, which Rinkata cannot offer.

Jira

Jira is the default for larger engineering orgs, with a Free plan for up to 10 users and Standard at $7.91 per user per month. It has far more workflow controls than Rinkata, but none of its spec-to-proof tracing for agents.

Final Verdict: Is Rinkata Worth It?

Rinkata is aimed at a real and growing problem: agents that wander from what you asked for, lose context between sessions and report work as done without proof. The goal, spec, decision and proof chain, the human-only approvals and the checked screenshots are thoughtful answers. The legal pages are among the most candid we have read, down to naming which features send data to which model.

It is still early. The jargon is heavy, export and deletion need an email, data stays in one US region, and at $15 a seat it costs more than the trackers most teams already pay for. Our advice: start on the free plan with one project and one agent. If your agents stop inventing context within a couple of weeks, Team is worth it for a small crew. If they do not, stay with the tracker you have.