el-gringo 1.1.0

El Gringo - Multi-Agent AI Orchestration Platform

El Gringo

Multi-agent AI dev tool. Ask 5 AI models, get one answer.

---

Install

pip install el-gringo

Setup (60 seconds)

elgringo setup

The wizard walks you through adding API keys, detects local models, and writes ~/.elgringo/config.yml. Skip any provider you don't have.

No keys at all? Use local models for free:

ollama pull llama3.2:3b
elgringo setup    # select local only

Quick Start

# Ask anything
elgringo ask "explain this error: TypeError: 'NoneType' object is not subscriptable"

# Code review
elgringo review server.py

# Fix code (reads, plans, writes, tests)
elgringo fix "add error handling to the login endpoint" auth.py

# Generate tests
elgringo test utils.py

# Security audit
cat app.py | elgringo secure

# Autonomous build
elgringo build "a FastAPI REST API with JWT auth"

Claude Code Integration

elgringo mcp install
# Restart Claude Code — 73 MCP tools now available

Or check status:

elgringo mcp status

---

Commands (30+)

Code

Command	What it does
elgringo review <file>	Multi-agent code review
elgringo debug <error>	Debug an error
elgringo test <file>	Generate tests
elgringo refactor <file>	Refactor code
elgringo secure <file>	Security audit
elgringo optimize <file>	Performance tuning
elgringo explain <file>	Explain code

Coding (writes files)

Command	What it does
elgringo fix "task" [files]	Edit existing code
elgringo code "task" [files]	Write new code
elgringo implement "task" --commit	Implement + auto-commit
elgringo build "description"	Full autonomous build
elgringo undo	Revert last change
elgringo diff	Show last change

Business

Command	What it does
elgringo pitch "idea"	Investor-ready pitch
elgringo mvp "idea"	MVP spec + timeline
elgringo ship "product"	Launch checklist
elgringo market "product"	Marketing strategy
elgringo revenue "idea"	Revenue model
elgringo compete "vs X"	Competitive analysis

General

Command	What it does
elgringo ask "question"	Ask anything
elgringo write "topic"	Write content
elgringo analyze <file>	Analyze data/logs
elgringo research "topic"	Deep research
elgringo plan "goal"	Action plan
elgringo summarize <file>	Summarize content
elgringo deploy "task"	DevOps guidance

System

Command	What it does
elgringo setup	Configure API keys
elgringo mcp install	Add to Claude Code
elgringo stats	Cost breakdown + performance
elgringo models	Show all agents + status

Pipe support

cat error.log | elgringo debug
git diff | elgringo review
cat data.csv | elgringo analyze
cat README.md | elgringo summarize

---

How It Works

Your task
    |
+---+---+
| Router |  Picks the best agent(s) based on task type + past performance
+--+-+--+
   | |
+--+ +--+--------+
|GPT| |Gemini| |Grok|   Agents work in parallel
+--+ +------+ +----+
   | |    |
+--+-+----+--+
|  Consensus  |  Weighted synthesis (not concatenation)
+------+------+
       |
+------+------+
| Self-Correct |  If tests fail, retry with new strategy
+------+------+
       |
+------+------+
|   Memory    |  Store what worked, prevent repeat mistakes
+-------------+

The orchestrator:

1. Plans -- decomposes goals into dependency-aware subtasks
2. Routes -- picks the best agent per subtask (learned from feedback)
3. Executes -- runs subtasks in parallel batches
4. Validates -- sandboxes and tests generated code
5. Corrects -- retries failures with 7 adaptive strategies
6. Learns -- records outcomes so the next run is smarter

---

Free Tier

Works fully offline with Ollama. No API keys needed.

ollama pull llama3.2:3b
elgringo setup        # skip all cloud providers
elgringo ask "hello"  # runs locally

Cloud Agents

Add API keys for higher quality. Each provider adds one or more agents:

Provider	Agent	What it's good at
OpenAI	ChatGPT	Coding, debugging, optimization
Google	Gemini	Creative solutions, UI/UX
xAI	Grok	Reasoning, strategy, fast coding
Anthropic	Claude	Architecture, analysis, planning

---

Python API

from elgringo.orchestrator import AIDevTeam

team = AIDevTeam(project_name="my-app", enable_memory=True)

result = await team.collaborate("Design a database schema for a blog")
print(result.final_answer)
print(result.confidence_score)  # 0.0-1.0
print(result.participating_agents)  # ['chatgpt-coder', 'grok-reasoner']

REST API

uvicorn products.fred_api.server:app --port 8090

curl -X POST http://localhost:8090/v1/collaborate \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Review this auth system", "mode": "parallel"}'

---

Stats

• 69,000+ lines of Python across 149 modules
• 6 AI providers (ChatGPT, Gemini, Grok, Claude, Ollama, LlamaCloud)
• 8 collaboration modes
• 73 MCP tools
• 18 development tools
• 7 self-correction strategies
• 544+ tests passing

What This Is NOT

• Not a wrapper around one model -- orchestrates multiple models with routing, consensus, and memory
• Not a chatbot -- agents execute tools, write code, run tests, deploy
• Not autonomous without consent -- coding commands confirm before writing files
• Not a hosted service -- runs on your machine, your keys stay local

Security

See SECURITY.md for the security model, tool permissions, and vulnerability reporting.

License

Apache 2.0 - see LICENSE

Copyright 2024-2026 TheGringo AI