AI WORK,
EXECUTED WITH
PROOF.
C1 routes tasks across agents, verifies what actually completed, and settles outcomes only after proof.
PRODUCTS WE SHIPPED. PROOF INCLUDED.
SELO — Safety-first task runner for AI coding agents
Wraps any coding agent in a contained worktree, audits the diff, and signs every outcome. Open source, Apache-2.0.
Verify it yourself:
curl -s -o r.json https://raw.githubusercontent.com/C1-run/selo/main/samples/receipt-example.json selo verify r.json Receipt: c1f-1791161998185566000 Verdict: FAILED_SAFETY Hash: OK Signature: OK Result: VALID - forbidden file modified: README.md (matches: README.md)
Receipt from a recorded test run — the signature and hash checks are real. Run them yourself.
AI IS CHEAP.
WRONG EXECUTION IS
EXPENSIVE.
Traditional AI agents operate in best-effort loops. C1 enforces a deterministic state machine where every action is anchored to verifiable proof.
SEVEN STAGES. EVERY ONE LEAVES AN ARTIFACT.
No stage ends in a claim — every stage ends in an artifact a buyer can check. Stage five is the differentiator: nothing leaves the factory without a tamper-evident, Ed25519-signed Selo receipt.
ONE ENGINE. THREE PACKS. IN ORDER.
DATA ANALYSIS
Objective predicate: same input → same output. Where the factory proves itself first.
SECURITY
Adjacent to Selo itself. PoC-validated findings, no false-positive claims.
FINANCE
Highest revenue, heaviest compliance. Entered after the factory is proven.
YOU BUY THE OUTCOME. THE RECEIPT IS HOW YOU KNOW.
Every claim on this page links to a shipped product or a signed receipt. Building something that needs proof in the loop? Talk to us.