ReGrade®
Behavioral evidence for release decisions
Field-level analysis that catches regressions in API behavior by comparing real traffic across software versions.
How it works
Compare behavior before you approve a release.
ReGrade catches regressions in HTTP APIs by comparing a new build with recorded behavior from the current version. See changes in responses and latency before release.
The example advances every five seconds. Select a step to explore it.
1
A sensor acts as a proxy in front of your application and records real HTTP traffic. Capture QA sessions, developer testing or existing automation to build a reusable baseline.
2
Replay the traffic against a new build and compare every response with the recorded responses. Profiles normalize rotating IDs, tokens and timestamps so meaningful differences remain.
3
Review deltas (response differences) by severity and label them as known or dropped. ReGrade also compares latency overall and per endpoint. Inspect results in the web UI, through the API or with MCP. In CI, unexplained deltas or latency regressions can block a merge.
Illustrative example, not a live run
GET /api/orders/42
Capture the request and response.
A QA session passes through the sensor to the current application.
Recorded response
200 OK status: "paid" updated_at: "10:00:00"
Reusable baseline
The request, response and timing are saved for replay against another build.
Response time: 120 ms
Use Cases
QA recordings
Record manual or scripted QA sessions once, then replay them against every new build. No test scripts to write or maintain.
Dev testing
Record traffic against current code, make a change or refactor, then replay locally. Catch unintended changes in behavior or speed before you push.
CI testing
Replay recordings against each merge request's build. The pipeline fails on unlabeled deltas or latency regressions, so reviewers see exactly what changed.
CI enterprise dashboard
See every project's health from its latest merged change and a recent trend. Statuses are red, amber, stale, green or grey, so leads can spot unexplained regressions and projects that haven't been tested lately.
The same agent, the same bugs, a third of the time
We gave Claude Code the same codebase, the same two known bugs, and the same AI model. One session had ReGrade's structured deltas (response differences) to work from. The other had raw traffic replay. Here is what happened.
Ghost CMS v2.2.2 → v2.2.3 · Claude Opus 4.6 · 72 Requests · 3,328 Deltas
3.2×
Time Saved
3m 41s vs 11m 56s
1.8×
Cost Reduction
$2.02 vs $3.57
3.5×
Token Efficiency
1.2M vs 4.1M tokens
96%
Root-Cause Quality
of deltas traced to root cause
Wall-Clock Time
Estimated Cost
Total Tokens
API Calls
Tool Calls
User Turns
With ReGrade's structured delta tools, Claude traced 96% of flagged deltas to a single root cause, using 71% fewer tokens and finishing 3.2× faster than raw replay.
Ghost CMS API benchmark: 72 requests, 3,328 deltas, 2 known bugs. Results describe this two-session comparison; evaluate performance and cost on your own workload.
See it in action
How Curtail's ReGrade can help you
1. Self-Healing Code
Self-Healing Code with the ReGrade MCP server
ReGrade replays recorded traffic against the code your agents write and reports what changed. Agents can use the findings to attempt repairs, and your team reviews the result before merge.
- ✓Less spent per bug fixed
- ✓Fewer review cycles per change
- ✓Fewer unasked-for changes reaching review
- ✓Fewer vulnerabilities introduced
2. Merge Request
Automatic Merge Request Analysis
ReGrade replays recorded production traffic against every merge request, automatically surfacing behavioral and performance differences before code reaches production.
- ✓Faster, more confident reviews
- ✓Auditable regression artifacts
- ✓More insight than pass/fail test results
- ✓Behavioral evidence for investigating differences
3. Refactoring
LLM Language Refactoring & ReGrade
When your team migrates code between languages, ReGrade compares responses to recorded requests across versions. Review the observed differences before approving the migration. Memory-safety validation remains a separate step.
- ✓Behavior checks during migration
- ✓Automated equivalence testing
- ✓Confident language migrations
- ✓Checks on every iteration
Evaluate one service before expanding
Choose a baseline and candidate build, then record and replay representative requests in an isolated test environment. Have your team review the differences before deciding which findings should block a merge or release. Replayed traffic cannot cover paths you did not exercise or identify every vulnerability.
Measure repair results and review effort alongside LLM API spend, subscription fees and replay infrastructure. Agree on success criteria before requiring the workflow across your organization.
Hosted or in your infrastructure
Use the hosted ReGrade service, or discuss deployment in your own data center or private cloud. The recording proxy captures request headers and bodies as well as responses; captured data is uploaded to the storage used by your ReGrade deployment. Review sensitive-data handling before recording production traffic.
Access checks use organization and team permissions. Authorized users can request recording deletion. Discuss retention, redaction requirements and deployment controls with us before an evaluation; recording and agent integration require setup.
See the agent connection and recording setupWe will show you what ReGrade finds on a service you already run. Or start free: 10 replays and 100 MB every month, no credit card and no expiration.