Build with AI

How it works

A model's memory of a minimum wage or a notice period is often a year out of date, and it never comes with a source. Juriset gives your agent the record instead:

• While it codes, the agent asks the MCP server for the rule in a place on a date and reads the answer with its official source.
• To write the integration, it reads llms.txt and the OpenAPI document, both written for models.
• At run time, your product calls the API, so values update when the law does, and each decision keeps the version ID it relied on.

Step by step

1. Add the MCP server to your AI tool (below). It works without a key, within the keyless limits.
2. Paste the project prompt into your agent's instructions file: CLAUDE.md, AGENTS.md or a Cursor rule.
3. Ask for what you want built, in plain words: "Show the overtime rules for the employee's state on the timesheet screen, with the source link."
4. When the feature needs a key, create one and put it in JURISET_API_KEY in your environment, not in the chat.
5. Review the change like any other: every value should come from a Juriset response, with its version ID kept.

One prompt for your project

Put this in the file your agent reads at the start of every session. It is short on purpose: the agent reads the details from llms.txt and the OpenAPI document when it needs them.

## Jurisdiction rules (Juriset)

This project gets employment and payroll rules (minimum wage, overtime, paid leave, termination notice, payroll contributions) from Juriset: https://api.juriset.dev/v1. Docs for agents: https://juriset.dev/llms.txt and https://api.juriset.dev/openapi.json.

- Never hardcode a wage, rate, threshold or notice period, and never answer one from memory. Look it up with the Juriset MCP tools while working, and call the API from the code.
- Never guess a slug: find jurisdictions with GET /v1/jurisdictions?q=<name>.
- Keep the version id (rv_…) and the citation URL of every rule the code acts on, and store them with the record the decision produced.
- Federal, state or provincial and local layers stay separate, as the API returns them. Follow the response's `resolution`; do not merge layers or compute a value for a case.
- Treat `review_pending`, `coverage.note` and `missing_layers` as things to show the user.
- Read the key from the JURISET_API_KEY environment variable. Never print it, log it or commit it.
- Juriset is informational, not legal advice; keep its disclaimer where we show its values.

• Claude Code reads CLAUDE.md; Codex and many other agents read AGENTS.md.
• In Cursor, save it as a project rule in .cursor/rules/ and set it to apply always.

Set up your AI tool

Claude Code

claude mcp add juriset -- uvx juriset-mcp

It needs uv. To use the hosted server instead, with nothing to install:

claude mcp add --transport http juriset https://mcp.juriset.dev/mcp

Codex

codex mcp add juriset -- uvx juriset-mcp

Or add it to ~/.codex/config.toml by hand; the MCP guide has the lines.

Cursor

Add the server to .cursor/mcp.json (the MCP guide has the JSON), then turn it on under Cursor's MCP settings. Save the project prompt as an always-applied rule.

Windsurf, VS Code, Gemini CLI and other agents

Any agent that speaks MCP can run uvx juriset-mcp or connect to https://mcp.juriset.dev/mcp. The MCP guide has the config for each one. Agents without MCP can still call the REST API from code: point them at llms.txt.

Claude, ChatGPT and other chat apps

Where your plan allows custom connectors, add one with the URL https://mcp.juriset.dev/mcp. Chat apps connect without a key, so they get today's rules within the keyless limits; history and as_of need a key, which only a coding agent or your own code can send.

Your API key

• You do not need one to start: without a key you get today's rules, 60 requests an hour and 500 a day per IP address.
• With a key, the local server reads it from JURISET_API_KEY; the hosted server takes it as Authorization: Bearer. Your plan's limits and features then apply.
• Each key can be revoked on its own, so give each project or environment its own key.

Tips for good results

• Ask for the source. "What is the overtime rule in Washington? Cite it." gets the excerpt and the link, which is what you want to check.
• Give the date. Rules change on set days. Say "on 1 January 2027" rather than "next year".
• Keep the levels apart. If an answer merges federal and state rules into one number, ask the agent to show each layer and the response's resolution.
• Read the gaps. A coverage note (for example, city minimum wages are not covered) belongs in your product's UI, not in a code comment.
• Test with recorded responses, not with values typed into the test, so the tests stay true when the rules change.

Use with AI

Paste this into Claude Code, Codex, Cursor or any coding agent. It points the agent at this guide and at the machine-readable docs, so it works from Juriset's own descriptions rather than from memory.

Read https://juriset.dev/docs/build-with-ai and https://juriset.dev/llms.txt.
Then set this project up to use Juriset:
1. Add the Juriset MCP server to your own configuration (local: uvx juriset-mcp; hosted: https://mcp.juriset.dev/mcp).
2. Add the guide's project prompt to this repository's agent instructions file.
3. Find every hardcoded wage, rate, threshold or notice period in the code and list them, with where each is used, so we can replace them with Juriset lookups. Do not change them yet.
Never ask me for the API key in chat; read it from JURISET_API_KEY when needed.