Tiden is live on Product Hunt. See the launch.

Intent verificationfor coding agents

Your agents write the code.
Tiden verifies it does what you actually asked for.
It keeps your specs versioned beside the code.

Start free trial
$ curl -fsSL https://tiden.ai/install.sh | bash
From the team behind Qase
Heavy Swells at the West Coast of Jutland, Niels Skovgaard, 1894
QuillbaseQuillbase AppOverview

Quillbase App QB

Documents, permissions, and the APIs partner teams build on.

main · 205 specs · 1 open branch

Shippable · gate passing on main

Quality gate

Pass

All gate subjects passing

Latest run

Run #11 passed

7 of 7 · intent branch · now

Coverage

100% verified

205 of 205 specs · 0 without tests

Needs attention 1

intent/2026-09-23-scim-group-role-mapping awaits review

✓ Pass✓ intent✓ 7 tests✓ run #111 cycle

Review
QuillbaseQuillbase AppBranchesscim-group-role-mapping

intent/2026-09-23-scim-group-role-mapping

Captured intent · +1 spec · +7 tests · run #11 · Claude Code

Merged by Maya Okafor

Spec changes · QB-207 SCIM group to role mapping

  1. 3+An IdP group maps to one workspace role
  2. 4+A user removed from the group loses the role within the sync window
  3. 5+A changed mapping re-provisions the affected users

Proven by

  • QB-1056Mapping an IdP group to the admin role grants admin accesspassed
  • QB-1057A membership change re-provisions affected users within the sync windowpassed
  • QB-1058A user removed from the group loses the rolepassed
  • QB-1061An unmapped group grants no rolepassed
Ask for a testAccept the risk✓ Merged

Activity

  • CCClaude Code opened the branch from session a0bda64c6m ago
  • TTiden run #11 passed, 7 of 7 tests2m ago
  • MOMaya Okafor merged the branch into mainnow
QuillbaseQuillbase AppReleases2.6.0

2.6.0 prod

13 components · released 14:02 UTC from main

✓ Pass

Coverage

205 of 205

specs verified by a test that ran

Components

13 of 13

healthy at release

Production issues

0 open

in scope for this release

  • Web App64 tests✓ Pass
  • Public API41 tests✓ Pass
  • Billing28 tests✓ Pass
  • Auth33 tests✓ Pass
  • Reporters20 tests✓ Pass

Snapshot frozen at release · 14:02 UTC

Works besideClaude CodeCursorCopilotCodexGemini CLI

From the team behind Qase

Coding speed went up.
Delivery failed to keep pace.

  • Review became the bottleneck.

    Agents open pull requests faster than anyone can read them. One gaming publisher saw pull requests rise 4 to 5x while features shipped stayed flat.

  • The agent grades its own work.

    It writes the code, the tests, and the docs from one reading of the ask. If it misread the ask, every check still comes back green.

  • The ask is lost when the session ends.

    What you asked for lives in a chat that closes. The next session treats the last one’s docs as fact.

How Tiden works

Your agent writes the code and the tests. Tiden runs beside it.
A person resolves anything the loop can’t, and merges.

Tiden maps what your product does.

Tiden reads your code and writes down each spec.
It marks the ones no test proves.

Low Tide at Pourville, near Dieppe, Claude Monet, 1882
your product · 6 components · 17 specs
Proven by a testNo test yetTIDEN-377 Billing tier caps volume · no test

Each change is proven by a test.

Your agent starts each session from a working contract.
The session can’t close over unverified work.

Skagen Beach in Moonlight, P.S. Krøyer, 1899
intent/rate-limit-public-api
Add rate limiting to the public APIClaude Code
  1. Started
  2. Coding
  3. Settle
  4. Close

Promises

  • Free returns 429 above 60 requests a minuterunning2 tests passed
  • Every 429 carries Retry-After in secondsrunning1 test passed
  • Pro returns 429 above 600no test

Can’t close yet: 1 spec has no test.

The spec changes with the code.

Green tests only prove the agent agreed with itself.
Tiden drafts the update on every merge, even ones that skip Tiden.

Grey Day by the North Sea, with Surf, Thorvald Niss, 1890
rate-limit-public-api
Gitwhat changed
api/public/v1/limiter.go
41func (l *Limiter) Allow(
42 ctx context.Context,
43 key WorkspaceKey,
44) bool {
45 q := l.tiers.For(key)
46 return l.win.Count(key) < q
47}
41// TODO: rate limits
Tidenwhat it was supposed to do
TIDEN-377 Billing tier caps volume
3Free returns 429 above
4 60 requests per minute.
5Pro returns 429 above 600.
6Every 429 carries
7 Retry-After in seconds.
3Unlimited for all tiers.

Review what was asked and proven.

Each branch shows what changed and what proves it.
Code review stays with your reviewer.

A Beach with Dunes. The West Coast of Jutland, Dankvart Dreyer, 1843
intent/rate-limit-public-api → main
✓ Pass3 of 3 promises delivered · 0 open
  • specTIDEN-377 Billing tier caps volume3 lines
  • + testfree_tier_returns_429_above_60runningpassed
  • + testretry_after_header_in_secondsrunningpassed
  • + testpro_tier_returns_429_above_600runningpassed

One verdict per release, part by part.

Errors in production point back to the release.
Your agent fixes them and adds the missing test.

Evening at Skagen, P.S. Krøyer, 1893
releases / 2.6.0
2.6.0prod✓ Pass
Coverage205 of 205
Components12 of 12
  • Web App✓ Passrisk 15/28
  • Public API✓ Passrisk 14/25
  • Billing✓ Passrisk 12/22
FeatureFlagWarningSearch · first seen in 2.5.0 · no test covered itFix with agent

Connects to the tools your team already uses

  • GitHub
  • GitLab
  • Jira
  • Linear
  • Notion
  • Azure DevOps

FAQ

More answers in the docs.

My agent already writes and runs its own tests. Why do I need Tiden?

Its tests come from the same reading of the task as its code. If it misread what you asked, the tests pass anyway. Tiden checks the change against the intent captured before the session started, and your existing tests catch the regressions its new tests never mention.

How is Tiden different from spec-driven development?

Spec-driven tools like GitHub Spec Kit, OpenSpec, and Kiro help you write the spec before the agent starts coding. Tiden checks the code against it after the agent finishes coding, on every change and with any agent, and keeps the evidence. Already have specs in Markdown? Import them.

Can’t we build this with our own tests and agents?

You could. You’d be building intent capture for every session, specs that stay current as the code changes, tests mapped to each change, and evidence for every release. One QA automation lead spent a year on it before giving up. Tiden starts at $49 a month.

Do we need to write specs by hand?

No. Connect GitHub or GitLab. Tiden reads your code and your tickets in Jira, Linear, or Notion and drafts the specs. You review each draft on a branch before it merges.

Do we need tests before we start?

No. Tiden marks what no test proves, and your agent writes the missing tests. Already have a test suite? Import it. Tiden links each test to the feature it proves and keeps it current as the product changes.

Midsummer Night at Tisvilde Beach, Julius Paulsen, 1886

Ship agent code you can prove is right.

Connect GitHub. Tiden maps the rest.