Run Durable Agent Jobs. Prove What Happened.
Gait is an offline-first runtime that dispatches durable agent jobs, captures state-changing tool calls, and emits signed packs you can verify, diff, and turn into deterministic CI regressions.
# Install
curl -fsSL https://raw.githubusercontent.com/davidahmann/gait/main/scripts/install.sh | bash
# Create a signed pack from a synthetic agent run
gait demo
# Prove it's intact
gait verify run_demo
# Turn it into a CI regression gate
gait regress bootstrap --from run_demo --junit ./gait-out/junit.xmlDurable Jobs: Run Without Losing State
Dispatch multi-step, multi-hour agent work with checkpoints, pause/resume/cancel, approval gates, and deterministic stop reasons.
Signed Packs: Portable Proof
Every run and job emits a signed pack you can verify, diff, and inspect offline. Attach it to PRs, incidents, and audits.
Regress: Incident to CI Gate
One command converts a failure into a permanent regression test with JUnit output and stable exit codes.
Gate: Fail-Closed Policy Enforcement
Evaluate structured tool-call intent against policy before side effects execute. Non-allow means non-execute.
Voice Agent Gating
Gate high-stakes spoken commitments before they are uttered. Signed SayToken capability tokens and callpack artifacts for voice boundaries.
Context Evidence
Deterministic proof of what context the model was working from. Privacy-aware envelopes with fail-closed enforcement when evidence is missing.
Why Teams Adopt Gait
| Without Gait | With Gait | |
|---|---|---|
| Long-running agent work | fails mid-flight, lost state | durable jobs with checkpoints + resume |
| Incident evidence | logs + screenshots | signed pack + ticket footer |
| Regression loop | manual repro, often skipped | deterministic fixture + CI gate |
| High-risk tool calls | best-effort guardrails | fail-closed policy + approvals |
| Voice agent commitments | hope they say the right thing | gated before speech + signed callpack |
| Audit posture | incomplete reconstruction | offline verifiable signed artifacts |
Frequently Asked Questions
What problem does Gait solve for long-running agent work?
Multi-step and multi-hour agent jobs fail mid-flight, losing state and provenance. Gait dispatches durable jobs with checkpointed state, pause/resume/cancel, and deterministic stop reasons so work survives failures and stays auditable.
What does Gait do that logs do not?
Gait produces signed packs and traces with deterministic verification, so incidents are portable, independently verifiable evidence rather than best-effort log interpretation.
Does Gait require a hosted service?
No. Core workflows are offline-first and run locally: capture, verify, diff, policy evaluation, and regressions can run without a network dependency.
How does Gait handle prompt-injection style risk?
Gate evaluates structured tool-call intent at execution time and blocks or requires approval based on policy. Non-allow outcomes do not execute side effects.
Can Gait gate voice agent actions?
Yes. Voice mode gates high-stakes spoken commitments (refunds, quotes, eligibility) before they are uttered. A signed SayToken capability token must be present for gated speech, and every call produces a signed callpack artifact.
What is context evidence?
Context evidence is a deterministic proof of what context material the model was working from at decision time. Gait captures privacy-aware context envelopes and enforces fail-closed policy when evidence is missing for high-risk actions.
How do I turn a failed agent run into a CI gate?
Run gait regress bootstrap --from <run_id> --junit output.xml. This converts the run into a permanent regression fixture. Exit 0 means pass, exit 5 means the same drift was detected. Wire the JUnit output into any CI system.
Can I replay an agent run without re-executing real API calls?
Yes. gait run replay uses recorded results as deterministic stubs so you can debug safely. gait pack diff then shows exactly what changed between two runs, including context drift classification.
First pack in 60 seconds. Durable jobs, voice gating, and policy enforcement included.
Install, create a signed artifact, and turn it into a permanent CI gate — all offline.
Open Install GuideFor assistant and crawler discovery resources, use LLM Context.