Autonomous operator agent
The operator
that admits failure.
rig-captain runs plan → execute → verify, with a verifier that ignores the executor’s own success claim entirely — and when verification fails, it switches strategy instead of retrying the same plan twice.
rig-captain · plan-execute-verify loop · strategy-switch recovery[higgsfield: gold-lit ship’s wheel silhouette on a dark bridge, compass rose overlay, single spotlight, cinematic noir]
Why teams install
What rig-captain earns back
Never trusts claimed_success
verify() recomputes pass/fail from acceptance_criteria met — the executor’s own optimistic self-report is discarded entirely.
‹2c6301c›Recovers 4 of 5 structurally-trapped objectives
25% of the 20-objective benchmark trip the default strategy by design; strategy-switching recovers 80% of those.
‹50fed99›One unrecoverable objective, and it says so
1 of 20 objectives is structurally outside every configured strategy — rig-captain reports it unreachable, not fudged as a pass.
‹1c566fe›How it works
Three steps. No faith required.
01
Plan
Select a strategy — direct, decompose, tool-augmented — for the objective.
02
Execute
Run it. The executor reports claimed_success — and it is ignored.
03
Verify or Switch
Independent verification recomputes pass/fail; a fail rotates to the next strategy, never a blind retry.
Proof band
One packet. Every claim accounted for.
packet rig/rig-captain · ops loop v1 benchmark 20 objectives · 25% trap-designed · 3 strategies result recovery_rate 0.800 (4/5, ≥ 0.70) · final_success 0.950 (19/20, ≥ 0.85) gate independent_recompute cross-checked against report_metrics — both must agree anchor PASS — verifier walked every raw attempt from scratch ‹2c6301c› themis SEALED — merge gate armed
Every rig-captain run emits a packet like this — hash-chained, verifier-signed, and queryable from your own dashboard. How ProofPackets work →
Buyers
Who runs rig-captain
VP Engineering
autonomous-agent product
Agent Platform Lead
enterprise copilot
Founder
agentic ops tooling
Objections
Answered before you ask
Isn’t this just retry-with-backoff?
Retry-with-backoff repeats the same plan hoping for a different outcome. rig-captain rotates strategy — direct, decompose, tool-augmented — on every failed verification, and disabling that switch drops recovery to exactly 0%.
How do you know the executor isn’t just self-reporting success?
verify() never reads claimed_success. It recomputes pass/fail from acceptance_criteria met, independently, every time.
What happens to genuinely impossible objectives?
1 of 20 benchmark objectives is structurally unreachable by every configured strategy — rig-captain reports it unrecoverable instead of quietly inflating the success rate.
Pricing
Start free. Pay when it pays.
Generalist ships today. Domain Expert adds vertical memory + priors. Custom is built to your stack.
Free
Priced after a short conversation
5 objectives a month, single strategy.
Team
Priced after a short conversation
Unlimited objectives, full strategy repertoire, recovery dashboards.
Pro
Priced after a short conversation
Dedicated verifier fleet, custom strategy sets, SLA.
SOC2 Type II readiness targeted Q4 2026 — every recovery decision is packet-sealed and audit-exportable now.
