MCP server

The server runs on your machine over standard input and output (stdio). It’s read-only: it never writes files. When it suggests a fix, your coding agent shows you the change and applies it only after you approve.

How it tells your agent to behave

When your agent connects, Lintsure sends it instructions. In short, the agent should:

• never say or imply you’re compliant, certified, approved, audit-ready or “good to go”
• never make a decision for you, and ask before changing any file or treating a framework as applicable
• say that results cover code only, against Lintsure’s reading of the publicly available text of each framework
• always mention the work outside code, and recommend a qualified person reviews the results

Every tool reply also ends with the disclaimer, and a smoke test in the repository checks this on every build. Your coding agent is a separate product, so how closely it follows these instructions depends on the agent.

Tools

Tool	Inputs	Returns
scan	path (default .), optional frameworks	The Markdown report and the JSON report. Without frameworks, it uses lintsure.config.yaml in that path, or all available frameworks.
classify	Answers about your product: euUsers, personalData, specialCategoryData, usesAI, generativeAI, interactsWithPeople, decisionArea, b2bCustomers, wantsCertification, usMarket	Suggested frameworks with a reason for each, and a reminder to confirm with you first.
explain_framework	framework id	Who it applies to, key dates, each obligation marked as checked in code, partly in code or outside code, and common mistakes.
explain_rule	ruleId	The rule’s title, the controls it maps to, the fix and the auditor note.

decisionArea takes one of: none, credit, employment, education, essential-services, insurance, biometrics, law-enforcement, migration, justice, critical-infrastructure.

Set it up

The server runs with npx -y @lintsure/mcp. The examples below use that command. If you build from source instead (see Quickstart), use node /path/to/lintsure/packages/mcp/dist/src/index.js.

Claude Code

claude mcp add lintsure -- npx -y @lintsure/mcp

Cursor

In .cursor/mcp.json:

{ "mcpServers": { "lintsure": { "command": "npx", "args": ["-y", "@lintsure/mcp"] } } }

Codex

In ~/.codex/config.toml:

[mcp_servers.lintsure]
command = "npx"
args = ["-y", "@lintsure/mcp"]

Windsurf

In ~/.codeium/windsurf/mcp_config.json:

{ "mcpServers": { "lintsure": { "command": "npx", "args": ["-y", "@lintsure/mcp"] } } }

Each client’s settings format can change between versions. If a snippet doesn’t work, check your client’s own MCP documentation for where server entries go.

Try it

Ask your agent:

• check this repo for compliance
• which frameworks apply to us? we have EU users and use AI for credit decisions
• explain what the EU AI Act asks of our code