AgentClash

CI/CD

CI/CD regression gates for AI agents

Model, prompt, RAG, and tool changes should not ship on vibes. AgentClash runs repeatable agent workloads in CI, compares candidates to baselines, and blocks merges when behavior regresses.

live eval
gate: pass

Candidate

92correct patch, low cost

Baseline

88stable reference run

Control

73missed edge case

replay timeline

1loaded task inputs and tool policy
2ran sandbox actions and captured artifacts
3scored trajectory and validator evidence
4attached scorecard and release verdict

ci verdict

Candidate clears release gate

Correctness improved, latency within budget, and required artifacts were preserved for review.

agentclash run create --follow

What CI/CD agent eval needs

Built for reviewable agent decisions

Release engineering needs pinned datasets, stable scoring, and enough replay evidence to debug a failed gate without reproducing the issue manually.

OpenTelemetry-compatible trace import

Pinned datasets and golden test cases

Baseline versus candidate regression checks

Replay trails for tool calls, outputs, and artifacts

Scorecards for correctness, cost, latency, and evidence

CI gates for prompt, model, RAG, and tool changes

Workflow

CI gate workflow

Import the evidence

Start from OpenTelemetry traces, curated datasets, support transcripts, or a real failure your team already saw.

Pin the baseline

Record the current accepted behavior so every prompt, model, RAG, or tool change has a fair comparison point.

Replay the evidence

Inspect tool calls, outputs, artifacts, latency, cost, and judge evidence when a candidate gets worse.

Gate the release

Compare candidate and baseline runs, then fail CI before a regression reaches users.

FAQ

CI/CD agent evaluation FAQ

How do agent eval gates fit into CI/CD?

A dataset or challenge pack runs on every candidate change, AgentClash compares the scorecard to a baseline, and the pipeline fails when configured thresholds regress.

What happens when a gate fails?

Reviewers open replay evidence, inspect tool calls and artifacts, and either fix the regression or update the baseline intentionally.

Can gates cover cost and latency budgets?

Yes. Scorecards include correctness and evidence quality plus cost and latency signals you can enforce in release policy.