ECA - Free Open Source AI Pair Programming in Any Editor - ECA

One Config To LLM Them All

Everything is driven by a single JSON config. Here are a few things you can do.

🌐 Any Provider 🎯 Skills 🪝 Hooks 🔗 MCP Servers Custom 🤖 Agents ⌨️ Commands 🔧 Tools

Providers

Use any LLM provider

Built-in support for OpenAI, Anthropic, Copilot, Gemini, and Ollama. Point to any OpenAI-compatible API — DeepSeek, LM Studio, OpenRouter, your company’s internal endpoint — with a few lines of config.

Explore models config

~/.config/eca/config.json

{
  "providers": {
    "my-company": {
      "api": "openai-chat",
      "url": "${env:MY_COMPANY_API_URL}",
      "key": "${env:MY_COMPANY_API_KEY}",
      "models": {
        "gpt-5": {},
        "deepseek-r1": {}
      }
    }
  }
}

Skills

Teach the LLM new abilities

Skills are structured knowledge units the agent loads on demand. Bundle instructions, scripts, and assets into a folder — the LLM learns to use them when the task matches. Follows the agentskills.io spec.

Explore skills config

~/.config/eca/skills/lint-fix/SKILL.md

---
name: lint-fix
description: Learn how to lint and fix code
---

# Instructions

Run `clojure-lsp diagnostics` to find issues.
For each diagnostic, apply the suggested fix.
Always run diagnostics again after fixing to
confirm no regressions.

Hooks

Automate before & after events

Intercept tool calls to enforce policies, get desktop notifications when prompts finish, or run validation scripts automatically. Wire any shell command to any ECA event.

Explore hooks config

~/.config/eca/config.json

{
  "hooks": {
    "notify-me": {
      "type": "postRequest",
      "visible": false,
      "actions": [{
        "type": "shell",
        "shell": "notify-send 'Prompt finished!'"
      }]
    }
  }
}

MCP Servers

Plug in any MCP server

Connect MCP-compatible servers for memory, databases, APIs, or anything else. Tools, resources, and prompts from the server become automatically available to the agent.

Explore MCP config

~/.config/eca/config.json

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-memory"]
    },
    "postgres": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres",
              "${env:DATABASE_URL}"]
    }
  }
}

Custom Agents

Build specialized AI teammates

Create agents and subagents with their own system prompts, models, and tool access. Subagents run in isolated contexts for parallel work without polluting the main conversation.

Explore agents config

~/.config/eca/config.json

{
  "agent": {
    "reviewer": {
      "mode": "subagent",
      "description": "Code review specialist",
      "systemPrompt": "You are a senior code reviewer...",
      "defaultModel": "anthropic/sonnet-4.5",
      "maxSteps": 25
    }
  }
}

Custom Commands

Your own slash commands

Drop a markdown file in .eca/commands/ and it becomes a slash command. Use $ARG1, $ARG2 for arguments — perfect for repeatable prompts, review checklists, or team workflows.

Explore commands config

~/.config/eca/commands/check-performance.md

# Available as /check-performance

Analyze $ARG1 for performance issues.

Focus on:
- N+1 queries and unnecessary DB calls
- Memory allocations in hot paths
- Blocking I/O on async threads

Suggest fixes with benchmarks when possible.
If no file is specified, scan the most recent
git changes.

Custom Tools

Give the LLM any CLI tool

Wrap any command-line program as a tool the agent can call. Define the schema, the command template, and ECA handles the rest — the LLM learns to use it automatically.

Explore tools config

~/.config/eca/config.json

{
  "customTools": {
    "web-search": {
      "description": "Fetches URL content as Markdown.",
      "command": "trafilatura -u ",
      "schema": {
        "properties": {
          "url": { "type": "string" }
        }
      }
    }
  }
}