Your agents repeat work your team already did

FML captures coding-session history and automatically supplies relevant prior sessions and file provenance while your agents work.

Get started
~/app
$ curl -fsSL https://fml.inc/install.sh | bash
✓ session capture ready
✓ local trace store
✓ MCP query available
● context injection — prior sessions and file provenance as you work

The same task, with and without history

Prior work and the dead ends load before the agent starts.

Without FML

~/app — codex
> pin the dashboard to 30 days for now
Reading 14 dashboard components…
Trying a period-prop shortcut…
Same shortcut was tried and reverted last month — not visible to this agent
Rediscovering the selected-window dependency…

118k tokens · illustrative

With FML

~/app — codex
> pin the dashboard to 30 days for now
● FML context injected
8fe325a reverted this exact shortcut
Selected-window reads were the real dependency
Keep the window switcher; fix the data path

31k tokens · illustrative

Same task, same repo. Token totals are illustrative until published from a measured session pair.

What engineers get

FML changes what the agent knows at the start of a run. Nothing changes about how you work, and context only comes from repos you already have access to.

Known dead ends

Failed approaches appear before implementation begins.

Handoffs

The prior session comes along when work moves between people or machines.

Provenance

Every injected fact links back to the session that produced it.

Status

Ask what shipped and what remains. There is no report to write first.

Getting started

~/your-repo
$ curl -fsSL https://fml.inc/install.sh | bash
✓ FML CLI installed
→ opening browser…
✓ integrations installed and signed in to acme-eng
✓ session capture active
Ready. Run your agents as usual.

Install once.

01

Install FML wherever your team codes. Claude Code, Codex, and Gemini sessions land in a local trace store, with optional team sync.

FML dashboard
updated 2m ago

Sessions

128

last 30 days

Spend

$3.8k

model usage

Recent sessions

Checkout retry fix

Claude Code · $4.18 · shipped

Auth migration notes

Codex · reused as context

Start capturing team usage

02

As engineers work, FML shows sessions, spend, stuck runs, reusable workflows, and the context graph forming across the team.

SessionsSpendWorkflowsContext graph
FML · query
> Recover the session I started on my laptop
Codex · 38m ago · 4 prompts
Last edit: app/login/page.tsx
Last prompt: “still overflows on iOS”

Ask FML from anywhere.

03

Use MCP, Slack, terminal, or the coding agent to recover prior work and inject context into the next run.

What you can ask FML

Ask FML to explain prior work, assemble task context, and show which setup patterns are worth reusing. Example output shown with illustrative data.

FML · MCP
MCP request checkout context
scope: local traces · team graph · last 30d
Evidence found
5 PRs · 17 sessions · 3 review threads
PR #398 reverted client retries
Constraint: preserve PCI boundaries
Files: payments.md, checkout.spec.ts
Recommended context
worker retry summary · checkout-review hook
Injected before the next agent run
Claude Code · MCP
> what should the agent know before
> changing checkout?
→ owners: @payments-team
→ last changed: PR #412 (3 days ago)
→ reverted path: PR #398
→ constraint: preserve PCI boundaries
→ session summary: worker retry migration
→ files: payments.md, checkout.spec.ts
→ recommended hook: checkout-review
Slack
@fml what should we reuse?
Ranked setup opportunities
1. checkout-review hook
used by 4 people · protects 12 sessions
2. api-contract skill
used by platform team · 0 failed reviews
3. permission rule for prod secrets
used by 2 people · org policy candidate
Preview adoption diff? [y/n]

Works with your stack

Coding agents

Captures sessions from Claude Code, Codex, Gemini CLI, and the other harnesses your team already runs.

Compression & caching

Stack freely. They lower the price of each turn; FML removes the turns that repeat old work.

Analytics & dashboards

Reporting comes out of FML as well; the same data also goes back into the agents.

Routers & gateways

Coming soon: capture from LiteLLM, OpenRouter, and Cloudflare AI Gateway.

Estimate how much of your AI spend is avoidable

Enter last month's AI coding model bill from your OpenAI or Anthropic dashboard. FML classifies captured sessions and identifies the share spent re-answering questions your team already answered.

Monthly AI coding spend

$50,000

$1k$10k$100k$1M$10M+

Avoidable share

up to 40%

Monthly avoidable spend

$20,000

Annual avoidable spend

$240,000

Retrospective classification of captured sessions — spend where the model would have acted differently with the team's history. Across customers and models this runs up to 40%; our measured team came in at ~32%. Not a savings guarantee; the pilot measures realized results against your own baseline.

Pricing

Start locally. Add Pro when you want the team record: sessions, PRs, costs, and reusable context across every agent run.

Free
$0
Local AI session capture
Local SQLite session store
Claude Code, Codex, Gemini CLI
Session timelines and AI summaries
Local search and cost tracking
ProMost popular
$24/seat/month

Everything in Free, plus:

Bring your own API key
Build a context graph across the team
Track outcomes from harness to PR
Compare setups like skills, hooks, MCP, and models across the team
Deliver just-in-time context in Claude, Codex, Gemini, and others
Gather insights around token use and AI code
Talk to FML from Slack, MCP, and CLI
Enterprise

For orgs that need the full platform at scale.

Everything in Pro, plus:

Deploy setups and workflows across teams
SSO, roles, and access control
Custom integrations for the context graph
Export traces to S3 and other pipelines
Annotation, eval, and training dataset capture
Team onboarding and priority support