[ 1 / 7 ]

Docs that stay in sync.
Automatically.

Documint detects documentation drift the moment your code changes. AST-level analysis finds stale docs instantly. AI-generated patches keep everything in sync. MCP integration for AI agents. CI/CD checks for every PR.

INSTALL
>npx @documint/cli@latest init
AST diffingAI patch chainMCP-readyCI enforcement
Live previewrepo connected
documint scan --watch
$ index skeehn/documint
[OK] 312 symbols indexed across 24 files
[DRIFT] content/docs/api-reference.md
Reason: parameter signature changed in patch endpoint
[PATCH] generated fix in 4.2s
last scan4.2s
docs updated3 files
review modehuman-in-loop
[ 2 / 7 ]

How AST-Powered Drift Detection Works

Three steps to keep docs in sync with code. Continuously.

01

Connect

Point Documint at your repo. One-time setup, no config files.

02

Watch

We diff the AST, not the prose. SHA-256 hashes of every signature.

03

Patch

4-step AI chain: diff, locate, write, verify. You approve and merge.

Code Push[+]
──→
AST Diff[x]
──→
AI Patch[*]
──→
PR Opened[+]
WORKS WITH YOUR TOOLS
GitHubVS CodeCursorClaude DesktopSlackGitHub Actions
[ 3 / 7 ]

Documentation Coverage, Accuracy & Automation

Symbol-level accuracy

We hash your AST, not your prose. No guessing which docs need updating.

Agent context files

Auto-generated .mint, CLAUDE.md, AGENTS.md, and llms.txt from live symbols.

Live sync + CI

Watch mode for local dev. CI integration for GitHub Actions.

[ 4 / 7 ]

Everything you need to keep docs alive

01SYMBOL-LEVEL DRIFT DETECTION
Your function gained a parameter. Your return type changed. Your class got a new method. Documint knows — because it hashes your AST, not your prose. Every push is diffed at the symbol level. When signatures change, the exact stale doc sections light up instantly.
- execute(input: string): Result + execute(input: string, retry: RetryPolicy): Result DRIFT: docs/api-reference.md § "Execute" is stale DRIFT: AGENTS.md § tool "execute" has wrong params
02AI PATCHES THAT VERIFY THEMSELVES
A fast model reads the diff and maps it to doc sections. A powerful model writes a surgical markdown patch — only the stale paragraphs, not the whole doc. A verifier checks the patch against the actual code. If it fails, it retries. You just approve and merge.
Analyze diffWrite patchSelf-verifyYou approve
03.MINT — YOUR CODEBASE AS A KNOWLEDGE GRAPH
Every function, every type, every relationship — extracted from your AST and packaged into a single .mint file. AI agents that read it understand your API at the symbol level. Not from a stale README. From the actual code, verified against the actual docs, refreshed on every push.
documint.mint (313 symbols, 24 files, 0.20s) ├─ SYM: full API graph with typed params + returns ├─ REF: exact file:line source references ├─ AIC: AI instructions (trust SYM over NAR) └─ exports: CLAUDE.md, AGENTS.md, llms.txt, API.md
04REAL-TIME WATCH MODE
Save a file. Documint parses the change in milliseconds, diffs the symbol table, and shows you exactly what drifted — the old signature, the new one, which doc section is now wrong, and the command to fix it. No waiting for CI. No context switching.
05CI THAT BLOCKS STALE DOCS
Add one line to your GitHub Actions. If docs are stale, the build fails. Every PR gets a summary table: which symbols changed, which docs drifted, severity level. Your team can never merge stale documentation again.
06LEARNS YOUR WRITING STYLE
Every patch you approve is stored. Documint learns how your team writes — the tone, the level of detail, the formatting conventions. By the tenth approved patch, the AI drafts in your voice. Powered by Voyage reranking over your approved patch history.
[ 5 / 7 ]

Install the CLI, MCP Server, or API

Connect your repo and run your first documentation drift scan in under 30 seconds.

npx @documint/cli@latest init
npx @documint/cli@latest generate
npx @documint/cli@latest status
[ 6 / 7 ]

Early access users

What developers say about Documint

Documint caught 8 stale API docs in our first week. No more "I thought we updated that".

Sarah C.
Platform Engineering Lead

The .mint files are game-changers for Claude. We can now give context to 50 endpoints in one request.

Marcus J.
Staff Engineer

We went from quarterly doc audits to real-time validation. This is how docs should work.

Elena R.
DevRel Lead
[ 7 / 7 ]

Pricing — Free Documentation Drift Detection to Start

PLAN

Developer

Free

For personal projects

  • 1 repository
  • 100 scans/month
  • Basic drift detection
  • Community support
Join waitlist →
PLAN

Pro

$49
/month

For small teams

  • Unlimited repos
  • Unlimited scans
  • AI patch generation
  • CI/CD integration
  • Slack notifications
  • Priority support
Join waitlist →
PLAN

Enterprise

Custom

For large organizations

  • Everything in Pro
  • SSO + SAML
  • On-premise option
  • Dedicated support
  • Custom integrations
Join waitlist →
EARLY ACCESS

Join the waitlist for early access

Be first to get symbol-level drift detection, AI patch generation, and .mint context files for your codebase.

Join waitlist →Sign up with GitHub · We approve manually · No spam
read the docs