Turn decisions into governed execution.

A control layer for turning data-driven decisions into governed actions.

Tessra turns intents into governed actions—with deterministic policy, approval gates, execution dispatch, and durable receipts.

Built to sit where decisions meet execution—directly alongside your data systems.

CALL APP.REQUEST_ACTION(
  action_key => 'ad_budget.increase',
  params     => { amount: 2500, campaign_id: 'c1' }
)
→ POLICY CHECK → APPROVED → EXECUTED → RECEIPT

A governed action — from any system that can make an HTTP call or run SQL

Request early access Watch demo

Already have access? Run your first action →

Runtime demo

Turning AI Actions Into Data

See Tessra turn an AI-requested budget change into a governed, queryable runtime object — with policy evaluation, approval, execution, and a durable receipt back in Snowflake.

Claude → Tessra → approval → execution → Snowflake receipt

Alternate 3-minute walkthrough (Snowflake → receipt) on YouTube

How it works

Request

An agent or SQL call proposes an action key and parameters.

Policy

Tessra evaluates deterministic rules against live warehouse data before anything runs.

Approval

If policy requires it, a human approves or rejects via Slack or email.

Execution

Once cleared, the action is dispatched to the configured external system.

Receipt

The outcome — verdict, context snapshot, approver — is written as a queryable row in your warehouse or database.

Tessra sits between callers and execution systems—policy and approvals run before dispatch; outcomes are recorded as receipts.

Why Tessra

Policy reads live data

Not static rules. At evaluation time, Tessra queries your warehouse. The ROAS that exists right now — not an hour ago — decides whether the action runs.

Receipts are just SQL

Every governed action becomes a queryable row. Use your BI tool, dbt, or any SQL client — wherever your data already lives.

Hard spend ceilings

Set daily or weekly budget limits per action type. Tessra auto-denies actions that would breach them. No prompt engineering required.

The decision — including the data at the moment of evaluation — as a permanent row in your warehouse:

SELECT
  action_key,
  params:amount::number        AS amount_usd,
  context_snapshot:roas::float AS roas_at_eval_time,
  verdict,
  approved_by,
  receipt_time
FROM APP.RECEIPTS_VIEW
WHERE action_key = 'ad_budget.increase'
ORDER BY receipt_time DESC;

Query receipts anywhere SQL runs — your BI tool, dbt, or any SQL client

Works across your AI stack.

Any system that produces a decision — an agent, a Cortex query, an automation workflow — can submit an intent to Tessra. Policy and approvals run before anything executes.

This is why agents cause incidents — not because they're unintelligent, but because the operational data that makes their actions consequential isn't consulted at the moment of execution. Should this ad budget increase fire? That depends on what ROAS is right now, in your warehouse. Tessra reads that data at evaluation time, before anything runs.

Describe a policy in plain English — Tessra generates and applies the governance SQL using the same warehouse context that evaluates it.

Watch runtime demo →Explore integrations →Policy authoring →

Start your first governed action in minutes

POST /v1/actions/request { action_key, params }

HTTP or SQL — policy, approval, receipt in minutes

Submit an action. Tessra governs how it runs—policy, approvals, execution, and receipts.

The proof lives in your data.

Request early access Watch demo