Findings are inputs.The fix is the outcome.
Aegis turns signals from the quality tools you already trust into a proposed patch with labeled evidence and cost — on demand, or on a schedule it keeps for you. You approve the change; Aegis writes it back and tracks whether it merged — or came back.
explicit approval · no auto-merge
Validate webhook signature before parsing
targeted fix · source-file write scope
- finding selectedstatic analysis · checkout.ts:84
- grouped into batch3 related findings · 1 draft PR
- cost estimated$0.42 est · $2.00 ceiling
- patch proposed2 files · +18 −5
- evidence attachedpartial · self-attested
Human approval required
no repository write has happened
the missing last mile
Great analyzers tell you what is wrong. Your team still has to turn the finding into a safe change — then learn whether that change stayed fixed.
one governed lifecycle
From selected finding to durable outcome
Aegis owns the handoff around your existing quality stack. The analyzer remains the authority on detection; the human remains the authority on repository change.
Bring the finding
Send coverage, static-analysis, dependency, duplication, or build signals from the CI and tools you already trust — or let scheduled runs and default-branch pushes surface them for you.
Get a proposed patch
A target-scoped run is claimed by an isolated per-client runner. It returns proposed files, a summary, and any submitted evidence.
Review before write-back
Inspect the diff, evidence label and completeness, and run cost. Reject it or explicitly approve the repository change.
Track the outcome
Aegis writes only the approved change to a branch or draft PR/MR, then records whether it merged or was later reverted.
more than a comment
The context to approve — and the record to learn
signals in
Keep the analyzers you trust
Aegis accepts coverage — line and branch — static-analysis, dependency, duplication, and build results instead of asking you to replace the tools that produced them.
always watching
Run on your cadence
Analysis re-runs on a schedule you set and on default-branch pushes, with next-due and scheduler liveness visible. Aegis watches so you don't have to remember.
targeted remediation
Fix one finding — or a batch
Dispatch a focused fix for a failing build, low-coverage file, or static finding — or group related findings into one batch that lands as a single draft PR.
economics
Cap the spend before dispatch
See the cost estimate and set a hard ceiling before a batch runs. Actual model cost stays attributable to the run and repository.
evidence
Know what backs the patch
Review submitted reproduction and regression summaries, their completeness, and an honest verification label before approval.
authority
Put a human on the boundary
The runner can propose files. It cannot write them back. A repository change requires explicit human approval — batches open as draft PRs.
outcome
Follow the fix past the PR
Aegis reconciles opened changes with provider state so merged and later-reverted outcomes stay visible.
fair grading
Grade only what you own
Generated and vendored paths are excluded from grading, and the exclusion list is disclosed and editable per repository.
keep the scanners
Aegis is the governed fix layer, not another analyzer fleet
Sonar-style platforms, security scanners, linters, and coverage tools are valuable finding producers. Aegis adds the review, approval, write-back, cost, and outcome lifecycle around the findings worth fixing.
Your tools
Detect + gate
Rules, locations, severity, coverage, and policy stay upstream.
Aegis
Propose + review
Targeted patch or batch, evidence label, estimated cost under a ceiling, and mandatory approval.
Provider
Merge + observe
Branch or PR/MR write-back with merge and revert reconciliation.
Evidence without the theater
Current runner evidence is shown as self-attested — not as Aegis-verified. Reviewers can see the submitted reproduction and regression summaries, evidence completeness, and run cost before deciding whether the patch may be written back.
Stop counting findings.Start closing the loop.
Connect a repository, keep your current analysis stack, and put a human-governed fix lifecycle — running on your cadence — behind the issues that matter.