Intelligence Packs
Request stack-specific analysis context from FinishKit for local code analysis where your code never leaves your machine.
Intelligence Packs are analysis bundles tailored to your technology stack. They contain framework-specific rules, security advisories, and analysis prompts that your AI editor uses to scan your code locally.
Why local analysis?
With Intelligence Packs, your source code never leaves your machine. FinishKit provides the analysis context; your AI does the analysis.
- Privacy: Your code stays local. Only stack metadata (framework, language, integrations) is sent to FinishKit.
- Speed: No waiting for a remote clone and scan. Analysis starts immediately after pack delivery.
- Offline iteration: Once you have a pack, you can re-analyze without network access.
How it works
Step 1: Request a pack
Your AI editor sends your stack details to FinishKit. This happens automatically when you use the request_intelligence_pack MCP tool, or manually via the SDK.
FinishKit detects your framework, language, package manager, integrations (Supabase, Stripe, etc.), and dependency versions.
Step 2: FinishKit synthesizes the pack
On the server, FinishKit builds a pack from its knowledge base:
- Framework-specific production readiness rules
- Known CVEs and security advisories for your dependency versions
- Community patterns and common mistakes for your stack
- Analysis prompts targeting all eight production readiness dimensions
Step 3: Your AI analyzes locally
Your AI editor receives the pack and uses it to analyze your code against the rules and prompts. Findings are generated locally with the same structure as cloud scan findings: category, severity, dimension, file reference, explanation, and suggested fix.
Step 4: Sync findings to your dashboard
Findings are uploaded to your FinishKit project via the sync_findings tool. They appear alongside any cloud scan findings in your Finish Plan, tracked and deduplicated.
Using Intelligence Packs
Via MCP
Ask your AI editor:
"Request a FinishKit Intelligence Pack for this project and analyze the code."
Behind the scenes, the AI calls request_intelligence_pack, analyzes your code with the pack context, then calls sync_findings to upload results.
Via SDK
const pack = await fk.intelligence.getPack({
framework: 'nextjs',
language: 'typescript',
packageManager: 'npm',
integrations: ['supabase', 'stripe'],
dependencies: { next: '16.0.0', react: '19.0.0' },
})Billing
| Plan | Packs per month |
|---|---|
| Free | 5 |
| Pro | Unlimited |
| Team | Unlimited |
Each pack request uses one credit. Syncing findings back to your dashboard is free.
Intelligence Pack credits and cloud scan credits are tracked separately. Using a pack does not count against your scan quota, and vice versa.