Runtime reliability for self-hosted agents

Your agent sees giants.

Quantized models don't fail loudly — they loop, fumble tool calls, and claim victories that never happened. Sancho rides alongside every run: it detects distress mid-episode and reroutes before the task fails. Run the cheap model. Keep the speed. Lose the failures.

$ pip install sancho && sancho serve
sancho: shadow mode — watching, not touching

▲ 14 distress events this week
  9 format-class   → recoverable by corrective prompt
  3 hard collapse  → would have escalated
  2 silent false-successes flagged

estimated recovered tasks: 8
$ sancho activate
✓ reroute ladder armed — zero code changes

Static metrics said fine. The agent said catastrophe.

We benchmarked the same model at different quantizations inside a deterministic agent harness — every verdict from checkers, never model self-report, every run reproducible from a manifest hash. Perplexity saw nothing. The agent loop saw this:

ConfigurationAgentic tasks cleanWhat happened
35B MoE · 2-bit quant, alone0 / 14Total collapse — loops, malformed tool calls
Same model · near-lossless, alone10 / 14Better — but it lied about one success
2-bit quant + Sancho15 / 15Distress caught mid-run, rerouted, recovered

Exploratory pilot data (N=1 per task), measured at a pinned runtime on two independent backends. Pre-registered, multi-seed confirmation in progress — every number on this page will reproduce with one command.

Judgment, riding alongside brilliance

An OpenAI-compatible proxy between your agent framework and your model servers. Change one base_url — no agent code changes.

I.

Detect

Malformed tool calls, degenerate loops, budget burn, logprob anomalies — runtime signals, scored for severity as the tokens stream.

II.

Diagnose

Format sloppiness or cognitive collapse? One cheap constrained turn classifies the failure — because the fix depends on the disease.

III.

Reroute

Corrective prompt first, precision climb next, model escalation last — the cheapest intervention that completes the task, chosen live.

The Sancho Reliability Index

Which models see giants? A public leaderboard scoring model × quant × runtime on agentic reliability — derail rate, silent-derail rate, clean rate — from deterministic, reproducible runs. Vendor-independent. Never pay-to-place. Launching with the preprint.

Get notified at launch

The squire becomes the governor

Everything a single box needs is free, forever. You pay when agents become something your business depends on across a fleet.

Squire
Free · Apache-2.0
  • Full detection + reroute ladder
  • Shadow mode + one-command activate
  • Flight Recorder + OTel export
  • Single node, your GPUs
Governor
Per GPU / month
  • The verifier — catches silent false-successes
  • Fleet policies + SLO reporting
  • Community-calibrated routing law
Barataria
Enterprise
  • Compliance pack (EU AI Act error-handling)
  • Air-gapped install + on-prem verifier
  • Private calibration on your workloads

Ride along early

Early access is small on purpose: self-hosted teams with agents in production get priority. One email when the repo, the Index, and the finding go live.

Request early access

Swap this button for your form endpoint (Buttondown / Formspree) before launch.