One install across every major AI agent

Install Euriklis as an MCP server

Add one config snippet to your agent and the full Euriklis op catalog becomes callable as tools. The agent declares a DAG, we validate and execute it.

Get an API key Read the docs

What is MCP?

The Model Context Protocol is an open standard for exposing tools to AI agents. An MCP server publishes a tool catalog; the agent discovers and calls those tools during a conversation. Euriklis ships a single MCP server that exposes the full op library as named tools.

Install

The server is distributed as @euriklis/mcp on npm. You only need an API key — set EURIKLIS_API_KEY in the agent's MCP config.

Claude Desktop

Open ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent on Windows / Linux, then add:

{
  "mcpServers": {
    "euriklis": {
      "command": "npx",
      "args": ["-y", "@euriklis/mcp"],
      "env": {
        "EURIKLIS_API_KEY": "eur_free_..."
      }
    }
  }
}

Cursor

Settings → Features → Model Context Protocol → Add server. Paste:

{
  "mcpServers": {
    "euriklis": {
      "command": "npx",
      "args": ["-y", "@euriklis/mcp"],
      "env": { "EURIKLIS_API_KEY": "eur_free_..." }
    }
  }
}

Claude Code

Add to your project or globally:

claude mcp add euriklis npx @euriklis/mcp \
  --env EURIKLIS_API_KEY=eur_free_...

Or add to .claude/settings.json in your project root:

{
  "mcpServers": {
    "euriklis": {
      "command": "npx",
      "args": ["-y", "@euriklis/mcp"],
      "env": { "EURIKLIS_API_KEY": "eur_free_..." }
    }
  }
}

Claude.ai

In a Project, open Settings → Tools → Add MCP server. Use the hosted endpoint: https://mcp.euriklis.aiPaste your API key when prompted. Works in Claude Pro and Team plans.

Codex

Codex consumes the same npx-launched MCP server. Add the Claude config snippet above to your Codex MCP settings file.

ChatGPT

ChatGPT Pro supports remote MCP endpoints. Use the hosted URL https://mcp.euriklis.ai with bearer-token auth using your API key.

Zed

Add to ~/.config/zed/settings.json:

{
  "context_servers": {
    "euriklis": {
      "command": { "path": "npx", "args": ["-y", "@euriklis/mcp"] },
      "settings": { "EURIKLIS_API_KEY": "eur_free_..." }
    }
  }
}

Available tools

Once installed, the agent sees these tools. The full op catalog lives in the documentation.

compute

Submit a DAG: declare inputs, list ops, name the output. Returns the materialised result.

compose

Combine two existing DAGs into one larger DAG and validate it without executing.

validate

Type-check a DAG without running it. Catches missing references, shape mismatches, cycles.

graph.algorithm

Run centrality, community detection, or shortest-path on a graph input.

tensor.op

Apply a single tensor primitive (norm, matmul, reduce) — the building blocks for compute.

optimize

Solve a least-squares, ridge, or constrained optimization stated as a DAG.

API key

Every install needs a key. The same key works against the HTTP endpoint:

curl https://api.euriklis.ai/v1/compute \
  -H "Authorization: Bearer $EURIKLIS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "inputs": { "Z": [[1,2],[3,4]] },
    "ops": [
      { "op": "normalize",     "x": "Z",  "as": "Zn" },
      { "op": "scalarProduct", "x": "Zn", "k": 3.14159, "as": "out" }
    ],
    "output": "out"
  }'

Free-tier keys are issued instantly on registration and include 1,000 calls per month.