Use Claspine with AI coding agents

Connect Claude Code, Cursor, Codex or Copilot to Claspine’s MCP server and give them Claspine’s rules, so they use the exact calls, fix errors and check their own work.

⏱ About 2 minutes

AI coding agents write most of the code that uses Claspine, so Claspine talks to them directly. The Claspine MCP server at https://api.claspine.com/mcp answers what runs where, the exact calls for a task, the API and Java types, what an error code means and how to fix it, and which upgrade steps apply; it also checks code and settings before a commit. It needs no account or key. The instruction files below put Claspine’s rules into your project for each agent. Code and settings sent to the check tools are analysed in memory and not stored.

Words you’ll see

API key: A long secret string that lets software use your account, like a password made for programs. Anyone who has it can spend on your account, so treat it like a password.

Steps

Claude Code: claude mcp add --transport http claspine https://api.claspine.com/mcp
Cursor: .cursor/mcp.json: { "mcpServers": { "claspine": { "url": "https://api.claspine.com/mcp" } } }
VS Code (Copilot): .vscode/mcp.json: { "servers": { "claspine": { "type": "http", "url": "https://api.claspine.com/mcp" } } }
Codex CLI: ~/.codex/config.toml: [mcp_servers.claspine] url = "https://api.claspine.com/mcp"
Other MCP clients: Streamable HTTP server at https://api.claspine.com/mcp; no authentication.
Codex, Cursor, Copilot, Gemini, Claude Code and any agent that reads AGENTS.md: add the rules with curl -fsSL https://claspine.com/agent/AGENTS.md -o AGENTS.md (or append the content to your existing file).
Claude Code (skill): add the rules with curl -fsSL https://claspine.com/agent/skills/claspine/SKILL.md -o .claude/skills/claspine/SKILL.md.
Cursor (project rule): add the rules with curl -fsSL https://claspine.com/agent/cursor/claspine.mdc -o .cursor/rules/claspine.mdc.
GitHub Copilot: add the rules with curl -fsSL https://claspine.com/agent/copilot-instructions.md -o .github/copilot-instructions.md (or append the content to your existing file).
Or set it all up in one command, without a token: npx @grahmur-org/claspine-cli init-agents writes the rules and the MCP server for Claude Code, Cursor and Copilot into this project, and npx @grahmur-org/claspine-cli doctor checks the project’s Claspine setup. Clients that only start local servers use npx @grahmur-org/claspine-cli mcp; MCP directories list Claspine as com.claspine/claspine.
Ask your agent something Claspine-specific, for example “add file uploads to this app with Claspine”. It should call what_runs_where and get_recipe, then run check_code on what it changed.
Optional, Claspine Atlas: let your agent set projects up for you. Add a second server, https://api.claspine.com/mcp/cloud (Claude Code: claude mcp add --transport http claspine-atlas https://api.claspine.com/mcp/cloud). The first time, a Claspine page opens: sign in, keep the permissions you want, choose the projects, and click Allow. The agent can then list projects, create free projects and API keys (with an expiry and an AI budget), set AI routes, and test storage and payments.
Remove an agent’s access any time on the console’s Account page, under Connected AI agents; it stops at once.

Claude Code, in your project: the MCP server and the Claspine skill.

claude mcp add --transport http claspine https://api.claspine.com/mcp
mkdir -p .claude/skills/claspine
curl -fsSL https://claspine.com/agent/skills/claspine/SKILL.md -o .claude/skills/claspine/SKILL.md

What goes where in Claspine

Console field	Value
Tools	what_runs_where, search_docs, get_guide, get_recipe, api_reference, java_api, explain_error, upgrade_notes, check_upgrade, check_code, check_config. All read-only.
Without MCP	https://claspine.com/llms.txt (summary) and https://claspine.com/llms-full.txt (every guide, recipe, error fix and upgrade note in one file).
Starter projects	Already contain an AGENTS.md for their stack.
What agents can never do on Claspine Atlas	Buy or change a plan, delete projects or the account, see stored provider keys or secrets, change live payment settings, download licence packages. For these the agent gives you the console page.

Keep it safe

The server never needs your Claspine keys. Never paste API keys, service keys or download tokens into an agent’s chat or MCP settings.
check_code redacts secrets it finds; still rotate any key that was committed.
A clean check is not proof of safety: keep code review and tests.
Claspine Atlas actions: only allow agents you just connected yourself; the consent page shows the agent’s self-chosen name and where it returns to. Every action is in the project’s audit trail with the agent’s name.

Claspine stores the values encrypted, never shows them again in full, and never sends them to your app or SDK.

If something goes wrong

The agent does not use the Claspine tools.: Mention Claspine in the request, or add the instruction file for your agent so it knows when to call them.
The MCP client says the server is unreachable or answers 405.: Use the Streamable HTTP transport with the URL https://api.claspine.com/mcp; the server answers POST only (no SSE stream).
Too many requests (429).: The server allows 120 requests per minute per address; wait a minute.

Provider documentation

← All setup guides