Correction Memory for AI Coding Agents — Now Available

Your agents keep making the same mistake. Prove the fix actually works.

FleetManifest captures the corrections your engineers make to Claude Code, Cursor, and Copilot, compiles the recurring ones into versioned rules, and measures whether each rule actually reduces the defect it targets — not just whether it shipped.

C
G
A
V
Trusted by engineering teams across the industry
fmanifest status
Live
23
Rules
14.2KB
Context/call
6
Never fired
3
Ineffective
prose
r_01H8K2db-raw-sql-concat
Retirement candidate
fires
skill
r_01H9P7missing-tenant-scope
Promoted 12 days ago
18 / 90d fires
check
r_01HB3Fauth-helper-bypass
Blocking in CI
4 / 90d fires
fmanifest analyse·4 defect classes uncovered

THE PROBLEM

AI agents make the same mistakes, repeatedly

Every organisation independently rediscovers the same 20–50 failure patterns specific to their stack. The correction lives in one of three places — and none of them work.

It lives in reviewers' heads

The senior engineer who knows the ORM call that deadlocks, the auth helper that gets bypassed, the migration convention agents ignore — that knowledge doesn’t scale, and it leaves when they do.

1.7x
more issues in AI-generated PRs than human PRs

It bloats an unmeasured CLAUDE.md

Hand-written instruction files are written once and never pruned. They grow monotonically, and because every line sits in the context window on every call, an unmaintained instruction file actively degrades agent performance over time.

+54%
more bugs per developer since agent adoption

It evaporates in the chat session

Tools that capture corrections and propose rule updates exist — but none of them measure whether the mistake stops happening. Without measurement, there’s no way to know which rules earn their context cost, so instruction files bloat and effectiveness decays.

31%
more PRs merge with zero human review

THE PLATFORM

Three layers. One measurable loop.

FleetManifest runs analysis on your own AI subscription, locally. Only abstracted, redacted summaries ever leave the machine — never your source code, prompts, or transcripts.

CAPTURE

Every correction, captured without slowing anyone down

Denied tool permissions, failed edits, agent giveups, and skill invocations are captured on the developer machine in under 50ms. Git forensics — reverts, human commits over agent branches, force-pushes — cover every agent, even ones with no hook surface at all.

  • PermissionDenied, PostToolUseFailure and StopFailure at <50ms p95
  • Git forensics fallback for Cursor, Codex CLI and Copilot
  • InstructionsLoaded gives ground-truth context inventory, not a disk scan
  • Hooks never call the network or an LLM — capture failure never breaks a session
$ fmanifest inventory --repo api-gateway
PermissionDeniedraw SQL string concat rejected[high]
PostToolUseFailuremigration ran against wrong schema[high]
bot_findingCodeRabbit: missing tenant scope[medium]
14.2KB always-loaded context · 6 rules never fired in 60 days

PROMOTION LADDER

Rules earn their place — or get demoted

Recurring defects start as prose. Once evidence shows a rule matters, it promotes to a skill, then to a deterministic check — each step trading context cost for enforcement strength and, critically, for observability. A rule that never fires and whose defect class has gone quiet gets demoted or retired automatically, as a proposal.

  • Promote prose → skill after 3+ occurrences, 30+ days distributed
  • Promote skill → check once a rule is deterministically checkable
  • Demote or retire on zero fires and zero recurrence
  • Flag as ineffective — never auto-act — when a rule fires but the defect keeps recurring
prose
CLAUDE.md / AGENTS.md
Context cost
Always loaded
Enforcement
Advisory
Firing observable?
No
skill
SKILL.md
Context cost
Metadata only
Enforcement
Advisory, scoped
Firing observable?
Yes — tool call
check
PreToolUse hook / lint rule
Context cost
Zero
Enforcement
Deterministic — blocks
Firing observable?
Yes — fires or not

MEASUREMENT

The one metric no competitor reports

Every yield chart shows the rate, the time series, the distribution marker and the sample size — never a claimed dollar figure or causal headline. Metrics roll up to repo, team, defect class or rule. Never to a named individual.

  • Recurrence rate — normalised by agent PR volume, never raw counts
  • Rule yield — recurrence before vs. after distribution, with a sample-size warning below 20 PRs
  • Unattended merge rate — the north-star metric, tracked over time
  • Confounder disclosure on every chart: model changes, team size, repo churn
missing-tenant-scope-on-queryRule distributed 46d ago
Before: 9.4% recurrenceAfter: 1.1% recurrencen = 74 agent PRs

HOW IT WORKS

From correction to measured rule

Analysis runs on your own AI, locally. Only its abstracted output — never source code, transcripts, or prompts — is ever transmitted.

01

Capture

Run fmanifest init in your repository. It installs the Claude Code plugin, scaffolds a CI workflow, and starts buffering high-precision signals locally: denied tool calls, failed edits, reverted commits, and findings from tools you already pay for.

02

Analyse & propose

On a schedule — not a chore someone has to remember — the analyser runs on your own AI subscription over the buffered signals and produces a proposal: new rules, promotions, demotions, retirements. Every change carries evidence with at least one occurrence id, or it never reaches the PR.

03

Distribute & measure

Merge the proposal PR and rules compile straight to CLAUDE.md, AGENTS.md, SKILL.md files, or hook checks. From there, recurrence is tracked per defect class, and the promotion ladder does its job automatically — proposed, never auto-applied.

A hard data boundary, not a guideline

A single redact() boundary validates every outbound payload against a strict allowlist. Unknown keys are a hard error. fmanifest sync --dry-run prints the exact JSON before anything transmits — and local-only mode sends nothing at all.

Allowed — abstracted only
  • Defect class ids, labels, descriptions
  • Rule tier, version, checksum
  • Event type, outcome, counts, durations
  • File extension, path depth, language
Never transmitted
  • Source code, diffs, file contents
  • Session transcripts, prompts, messages
  • Raw CI logs, stack traces, error text
  • Full file paths, branch names, commit messages

SOURCES & AGENTS

High-quality defect signal you already pay for

A large share of the best signal already exists in tools your team runs. Ingesting it costs nothing extra and gives labelled, structured defects from day one.

AI Coding Agents — capture depth varies, distribution is universal

C
Claude Code
Full hook capture
Cursor
Git forensics + .cursor/rules
GitHub Copilot
Git forensics + instructions.md
Codex CLI
Git forensics + AGENTS.md

Code hosting — SCM webhooks and attribution

GitHub
Code scanning + webhooks
GitLab
Vulnerability report API
Azure DevOps
Advanced Security alerts

Defect signal sources — findings you already generate

CodeRabbit
PR comment adapter
Semgrep
SARIF / code scanning
Sentry
Production error signal
Don't see your tool?
Rules distribute to any target that reads a plain instruction file or a hook — CLAUDE.md, AGENTS.md, .cursor/rules, Copilot instructions, or a Semgrep check.Read the docs →

THE PROMOTION LADDER

Rules aren't a flat list of prose

Every rule sits at one of three tiers, and moves between them on measured evidence — never automatically applied, always a reviewable proposal.

prose
CLAUDE.md / AGENTS.md
Context cost
Always loaded
Enforcement
Advisory
Firing observable?
No
skill
SKILL.md
Context cost
Metadata only
Enforcement
Advisory, scoped
Firing observable?
Yes — tool call
check
PreToolUse hook / lint rule
Context cost
Zero
Enforcement
Deterministic — blocks
Firing observable?
Yes — fires or not

A live rule registry, not a doc

prose
db-raw-sql-concat

Never construct SQL with string concatenation — use the query builder.

skill
missing-tenant-scope-on-query

Every query against a tenant-scoped table must apply the tenant filter.

18 / 90d
check
auth-helper-bypass

Blocks edits that reintroduce a direct call around the shared auth helper.

4 / 90d
skill
flaky-test-retry-pattern

Use the fixture factory instead of sleep-based waits in integration tests.

2 / 90d
bash — ~/projects/api-gateway
$fmanifest analyse && fmanifest propose
✓ 214 signals analysed, 3 defect classes evidenced
✓ Opened PR #482 — 1 promotion, 1 retirement, 1 new rule
$fmanifest apply --check
✓ rules.lock matches on-disk state — no drift
$fmanifest status
auth-helper-bypass check 4 fires / 90d
missing-tenant-scope skill 18 fires / 90d
legacy-retry-pattern skill 0 fires / 90d → retirement candidate
$_

PRICING

Flat pricing. Hard caps. No per-seat billing.

Every account starts with a 14-day trial of Team — no credit card — and converts into whichever tier fits, Solo included. Every self-serve tier is one flat price with a clear cap, so there's no usage meter and nothing to reconcile. Trial expiry and payment failure both degrade to local-only mode — they never break your CI.

Solo
$50/year
1 member · 3 repositories

The entry tier. One flat annual price — cheap, but paid: there is no free plan.

  • 1 member, 3 repositories
  • Full dashboard — Inventory, Rules, Outcomes
  • All source integrations
  • Annual billing only
Start 14-Day Free Trial
Most Popular
Team
$49/month
10 members · unlimited repositories

Flat rate for the whole team, whatever it does. The tier your 14-day trial runs on.

  • Up to 10 members
  • Unlimited repositories
  • Full dashboard — Inventory, Rules, Outcomes
  • All source integrations
Start 14-Day Free Trial
Growth
$199/month
40 members · unlimited repositories

Same flat shape, a larger cap. No per-seat billing and nothing to reconcile.

  • Up to 40 members
  • Unlimited repositories
  • Everything in Team
  • Priority source-adapter support
Start 14-Day Free Trial
Enterprise
Custom
Above 40 members

Defined by size, not by feature. Above 40 members the conversation is a sales one.

  • Unlimited members and repositories
  • Self-hosted docker-compose bundle
  • SAML 2.0 + OIDC SSO
  • Audit log export
  • Manual invoicing
Contact Us

Common questions

How does the 14-day free trial work?
You get 14 days of Team, no credit card required — and the clock starts when your backfill completes, not at signup, so setup never eats your trial. At the end you pick any tier that fits, Solo included; if you do nothing, the account degrades to local-only mode rather than locking you out.
What counts as a member?
A provisioned member of your organisation. Caps are enforced at invite time rather than billed for — going over the cap asks you to move up a tier, and you are never charged an overage. Repositories are only capped on Solo; every other tier is unlimited.
Is there a free plan?
No — but there is a free path. The plugin, CLI, and compiler are open source and run capture, analysis, and rule application locally, forever, for free. A subscription buys the hosted corpus: cross-repo distribution, recurrence measurement, the dashboard, and shared taxonomy.
What can 14 days actually show me?
The diagnostic, not the outcome. In two weeks you will see your context-cost inventory, never-fired and ineffective rules, defect classes surfaced from 90 days of history, and your first merged proposal. Recurrence decline and rule yield are defined over 90-day windows — those come with the subscription, not the trial.
Does FleetManifest see my source code?
No. Analysis runs on your own AI subscription, locally. Only abstracted, redacted summaries are ever transmitted — never source code, diffs, or transcripts.
What happens if payment fails?
Your account degrades to local-only mode rather than hard-locking. Capture, analysis, and rule application keep working; central sync and the dashboard pause.
Do I need to use Claude Code specifically?
No. Distribution and measurement are universal across Claude Code, Cursor, Copilot, and Codex CLI — capture depth is deepest on Claude Code today.

Your agents are already writing code. Prove your corrections are working.

Join engineering leaders who capture every correction, compile the recurring ones into versioned rules, and measure whether each one actually reduces the mistake it targets.

🔓14-day free trial, no credit card
📦Open source CLI
🔒Redacted by design — we never see your code
fmanifest init in under 5 minutes