A small team in Tallinn, building the inspection step for AI agents.
Suncly is built for the platform and security teams that have to say yes or no to an agent. We write the architecture down before the code, we keep the open questions visible, and we would rather ship a report that says 'not tested' than a score that hides it.
Three rules we apply to ourselves.
Approve less, never more.
Every fault in the system is designed to make Suncly approve less. Inconclusive is never a pass; no decision is never an approval.
Evidence over scores.
Counts per test case, transcripts, hashes and a signature. No universal score, because a score would hide what was not tested.
A human stays in the loop.
A person approves the test plan. A person resolves a flag. Suncly records who, and when.
Contact
- team@suncly.com
- Location
- Tallinn, Estonia
- Registration
- Company registration details will be published here.
Questions about Suncly.
What is Suncly?
Suncly is an evaluation tool for AI agents that speak the A2A (Agent2Agent) protocol. It reads an agent's Agent Card, tests every declared skill repeatedly in a sandbox, judges each run deterministically, signs the evidence, and reports what passed, what failed, what stayed inconclusive and what was never tested.
Who is Suncly for?
Platform and security teams that approve AI agents for use in their organisation and currently do that review by hand.
Is Suncly a model or an agent?
Neither. Suncly is a command-line evaluation tool that tests agents built on any model or framework through their A2A interface. In the current version it calls no model provider at all.
Where is Suncly based?
Tallinn, Estonia. Contact the team at team@suncly.com.
How is Suncly pronounced and spelled?
Sun-clee, one word, capital S: Suncly. The domain is suncly.com.
Run the first evaluation with us.
Suncly is in pilot. If your team approves A2A agents by hand today, tell us about one agent and one sandbox, and we will run the first evaluation together.
Or write to team@suncly.com.