Launching today
Tiden
Intent verification for coding agents
247 followers
Intent verification for coding agents
247 followers
Tiden verifies your coding agents built what you actually asked for. Most checks run inside the session that wrote the code, so the tests agree with it by construction. Tiden keeps your intent as requirements that branch and merge like code, and checks every change against that record. The verdict comes from tests run in your environment; no model grades the diff. You decide what merges, and production errors trace back to the release. Works beside Claude Code, Cursor, Codex and Copilot.
Hi Product Hunt, I'm Nikita, and this is Tiden, the product we've been building for the past year. It's for engineering teams shipping agent-written code every day who want to let agents do the coding while keeping control of what ships.
Qase is a software testing platform used by tens of thousands of builders, and 9 years of building it for them taught us a lot about how quality work gets done. Last year that changed on our own team: agents started producing most of the code, and the rest of our process didn't speed up with it. Review, testing, release, keeping requirements current: all of it became bottlenecks.
Other engineering teams tell the same story. A gaming publisher I talked to, with about 1k+ developers, saw pull request volume go up 4 to 5 times while the number of features they shipped stayed flat. A payments company described 10 times the coding output and 10% more delivery.
The problem underneath it
An agent session produces the code, the tests that pass on that code, and the docs describing what it built. The next session reads those docs as fact. Every artifact agrees with every other one, because they all came from the same place. A QA automation lead at a tax and accounting platform put it best: "If you have an agent that decides what to build, and then another agent is grading what's being built against what was actually built, you're essentially encoding the wrong thing. Your tests will pass, they'll be green, but it wasn’t what you wanted."
The two usual fixes, another rules file in the repo and better AI code review, both sit inside that same loop. The judge has to sit outside it.
What Tiden does
Tiden is that judge. It's intent verification for coding agents: it sits beside your agents, whichever ones you use, and verifies that what they built is what you actually asked for. It keeps that intent as specs, versioned like code, and checks every change against them rather than against the session that produced it.
Every agent session becomes a closed loop:
A session hook captures what you asked for before the first edit
The change comes back as a requirement for you to review; requirements branch and merge like code
Tiden picks the tests the change touches, runs them in your environment, and the agent fixes what fails before anything merges
The gate is tied to the spec. Red tells the agent what to fix. Green is a pass you can defend, with the evidence attached
In production, an error traces back to the release, the change and the prompt behind it, and the fix comes with the test that was missing
Tiden works beside Claude Code, Cursor, Codex and Copilot, and belongs to none of them. Specs live as one versioned record every agent reads from, instead of a growing pile of markdown. You set the thresholds and the gates, and a person approves anything the loop can't settle.
One limit, stated plainly: Tiden checks the work against what you asked for. Whether you asked for the right thing is still a human judgment.
Get started
Tiden is live at https://tiden.ai. The CLI installs in one command, the docs are at https://docs.tiden.ai, and Starter and Team both come with a 14-day free trial. If you'd rather talk to us first, there's a short form on the site: https://tiden.ai/contact.
What I'd like from you
Honest feedback over upvotes. Where does your team feel this gap most: review, regression, release, or production? What would make you trust an external check?
The real challenge with AI coding is not writing code faster but it’s making sure the code actually does what you intended.
@kshitij_mishra4 Yes, I’ve encountered exactly this problem many times. In fact, I noticed that the faster the code was written, the less confidence I had in it. I would spend hours verifying that the code addressed all of my original requests and had not broken anything else that might have been affected.
Tiden understands the blast radius from the outset. It tells the agent which parts of the application may be affected and asks it to run the existing tests for everything in scope. For new tests, it asks the agent to write them based on the customer’s original requirements, rather than on the generated code itself.
Can I use it without tests?
I just want to keep specs somewhere and use them across many repositories.
@anastasia_nikonorova Yes, it's possible!
Once you've generated the specifications within Tiden (this is part of the onboarding process), you can start sessions with agents. The lack of tests isn't a blocker here.
You can add tests later if you need them. Plus, Tiden will help you write them.
In the long run, it's best to have test coverage for requirements.
Engineer on the Tiden team here. My favourite bugs from building it were in our own gate, because they all had the same shape: it was too easy to pass.
Two of them:
A change that touched no known requirement passed. Nothing matched, so there was nothing to check, and "nothing to check" counted as green. Now it's a refusal, and accepting that risk is an explicit decision.
A requirement counted as verified on someone else's green run. Now only results from the change's own runs count.
Same lesson both times: a verifier is worth exactly as much as it is hard to fool. We build Tiden with Tiden, so we hit these ourselves first.
What's the sneakiest way your CI went green without really checking anything?
How easy is it to add Tiden to an existing project that's already using Cursor or Claude Code?
@maklyen_may Its straightforward create your product give your code repository to Tiden. Install CLI and Setup Tiden. Whole setup is done within half day and it starts working in background along with your coding agent Cursor, Claude Code etc. Its agent neutral by design. https://docs.tiden.ai/docs/getting-started/quickstart
@maklyen_may, we tried to keep it simple!
Once you install Tiden CLI (part of Tiden), it sets up all the skills and hooks you need.
After that, you can use Claude just like you normally would. The hooks and built-in Tiden skills handle everything behind the scenes.
Can Tiden catch cases where the code passes all tests but still doesn't match what the user asked for?
@daniel_henry4 Yes, that is USP of Tiden, it tests against the original intent captured as a spec even before single line of code is written by agent. So its verifies against intent - what user asked for. It even supplements missing impact that user may not have defined based on context graph which maintains all relationships and dependencies between different product feature.
Hey 👋 Another engineer from Tiden is here, too!
It's a really interesting experience to use Tiden to build Tiden. I think it could be interesting for other people too!
At first, changing so many domains at once seemed risky to me, but with Tiden, the process went smoothly.
Tiden helps track every detail and edge case while your product keeps changing, especially over the long term.
I believe you should give it a try!