API Reference

REST, and nothing surprising.

Predictable resources, standard verbs, cursor pagination. Rate limits below are the ones the server actually enforces, not aspirational figures.

Bearer tokens JSON only 4 MB request limit
The public /v1 surface is not open yet. The engine below is running today — roles, approvals, deployment, and the audit log are all enforced server-side. What is not finished is the stable public contract: versioned paths, an OpenAPI document, and the twelve-month deprecation guarantee. The shape documented here is fixed and will not change before release.

Authentication

Bearer tokens, workspace-scoped. A token inherits the role of whoever created it — an editor's token cannot approve, regardless of what it calls. Enforced server-side, not in the client.

Request
curl https://api.untactit.com/v1/assets \
  -H "Authorization: Bearer $TOKEN"

Sign-in itself is SSO only (SAML or OIDC). We never hold a password, so there is no password endpoint and nothing to leak.

Pagination

Cursor-based, not offset. Pass cursor from the previous response until has_more is false. Up to 100 items per page.

Response
{
  "data": [ /* up to 100 items */ ],
  "has_more": true,
  "next_cursor": "cur_01H8X..."
}

Resources

Six resources. Role requirements are enforced by the server and named in the error when you lack permission.

ResourceCoversMinimum role
assets List, read, create, update. Full version history per asset.viewer / editor
approvals Propose, approve, reject. Follows your workspace approval policy.editor / approver
deployments Plan, run, dry-run, and read deployment state including drift.approver
targets Connected platforms and their scope.admin
members Workspace members and role assignment.admin
audit Audit log with actor, action, and time filters. CSV export.admin

Errors

Machine-readable, and specific.

Errors carry a stable code you can branch on. When permission is the problem, the response names the role that would have worked — so you can fix it without guessing.

  • 403 insufficient role — includes required and actual
  • 413 payload over 4 MB
  • 429 rate limited
  • 409 conflict — version changed under you
403
{
  "error": {
    "code": "insufficient_role",
    "message": "Approving requires the approver role.",
    "required_role": "approver",
    "actual_role": "editor",
    "request_id": "req_01H8X..."
  }
}

Include request_id in support mail and we can find the exact call without asking you to reproduce it.


Rate limits

Token bucket per workspace, applied per path group. Sustained rate refills continuously; burst is what you can spend at once after idling.

300 / minute

Reads and asset writes. Burst of 50. Enough to sync a full inventory continuously and to save while editing without thinking about it.

30 / minute

Structural changes. Burst of 5. Role changes, workspace settings, and anything that alters who can do what.

12–120 / minute

Deployment, execution, external calls. Scaled to the cost of the work. A deployment is one call no matter how many machines it reaches.

Requests over 4 MB are rejected with 413. Enterprise workspaces with a concrete pattern can have limits raised — describe the workload and we will size it rather than quote a number that has nothing behind it.


API questions

When does the public API open?

No date we are willing to commit to publicly. The engine is running today; what remains is the stable contract and its deprecation guarantee. If API access is a requirement for your evaluation, tell us — that moves it up the list.

Is there a sandbox environment?

Not a separate one. Create a second workspace on the free plan and use it as your test environment — same behavior, no risk to production data.

How will you version it?

In the path. Breaking changes ship as a new version with twelve months of overlap on the old one. Additive fields appear without notice, so parse defensively.

Will there be an OpenAPI spec?

Yes, generated from the same source as the implementation so it cannot drift from actual behavior. A hand-maintained spec is a document that lies eventually.

Stop guessing what your agents are running.

Connect one workspace and see every skill, rule, and memory your team has in play — in about ten minutes.

Start free Talk to us

No credit card. Works with what you already run.