agentabi 0.3.0

Unified async Python API for driving coding agent CLIs — Claude Code, Codex, Gemini CLI, and more

agentabi

Unified interface layer for agentic coding CLIs.

One interface. Any coding agent.

What is agentabi?

agentabi provides a stable, unified interface (an "ABI") for interacting with different agentic coding CLIs. Write your integration once, swap agents with a config change.

Supported Agents

Agent	Provider	Status
Claude Code	Anthropic	Implemented
Codex	OpenAI	Implemented
Gemini CLI	Google	Implemented
OpenCode	Open Source	Implemented
Pi	Open Source	Implemented

Installation

pip install agentabi

With optional SDK integrations:

pip install agentabi[claude]   # Claude Code SDK support
pip install agentabi[codex]    # Codex SDK support
pip install agentabi[gemini]   # Gemini CLI SDK support
pip install agentabi[all]      # All optional SDKs

Note: Each agent's CLI must be installed separately (e.g., claude, codex, gemini, opencode, pi).

Quick Start

Run a task

import asyncio
from agentabi import Session

async def main():
    session = Session(agent="claude_code")
    result = await session.run(prompt="Fix the bug in auth.py")
    print(result["status"])       # "success"
    print(result["result_text"])  # agent's response

asyncio.run(main())

Stream events

async for event in session.stream(prompt="Explain this code"):
    if event["type"] == "message_delta":
        print(event["text"], end="")

Sync convenience

from agentabi import run_sync

result = run_sync(prompt="List Python files", agent="codex")

Discover available agents

from agentabi import detect_agents, get_agent_capabilities

agents = detect_agents()          # ["claude_code", "codex", ...]
caps = get_agent_capabilities("claude_code")
print(caps["supports_streaming"]) # True

Use Cases

• Fleet Management — Unified entry point for managing multiple coding agents
• Agent-to-Agent Calls — Translation layer for inter-agent invocation
• Benchmarking — Run the same task across agents, compare results
• Fallback & Routing — Automatic failover and cost-aware routing
• Middleware Pipeline — Inject logging, metering, security scanning, audit trails
• CI/CD Integration — Vendor-agnostic agent integration for pipelines

Ecosystem

agentabi is part of a layered stack:

agentabi  →  Agent CLI unified interface  →  like an OS ABI
llm-rosetta  →  LLM API format conversion    →  like a compiler IR

• llm-rosetta — LLM API format conversion layer for translating between LLM provider API formats (OpenAI, Anthropic, Google)

License

MIT