Don't trust.
Recompute.

Decentralized AI infrastructure anyone can provide, anyone can use, and everyone can verify — by recomputing the proof. Owned by no one.

Explore the standards

Meet trustless-ai.

Discover it

An open ecosystem of standards, reference implementations and tooling for AI agents you never have to trust — only recompute.

Recompute,
don't trust.

Every claim re-derived from the primary artifact, pinned to an exact reference, with the evidence attached. No summary taken on faith.

Owned by
no one.

Open standards and CC0 reference implementations — anyone can provide, anyone can use, no gatekeeper in the loop.

Verify
every proof.

A tri-state verdict — verified-good, verified-bad, or unverifiable. "Couldn't check" is its own answer, never a silent pass.

Verify a real proof →
Standards, tooling and reference impls — built in the open by independent contributors.
trustless-ai in practice

The stack

Three layers for agents you never have to trust — and you recompute at every one, from proof to deploy:

  • recompute-kit — verify. Re-derive any claim from its primary artifact, tri-state.
  • agent bindings — the ERCs encoding identity, provenance, verification and settlement.
  • onchain-boiler-kit — ship. Deploy the whole stack, no code required.

The recompute mesh

Independent nodes re-run every proof node-side, not self-attested — the always-on audit runtime for the agent-standards family. Verify by recomputing, never by trusting.

Know more
Meet the mascot

Take the agent for a spin

The trustless-ai mascot, yours to move. Let him run around — arrow keys or A/D to walk, Shift to run, Space to jump.

News & press

What's happening

Milestones and coverage — each card links straight to the artifact, so you can check it, not just read about it.