Client SDKs for the Agent Command Center

Python and TypeScript SDKs — plus LangChain, LlamaIndex, React, and Vercel AI SDK integrations — for Agent Command Center, Future AGI's open-source, OpenAI-compatible AI gateway.

Quickstart · Gateway repo · Docs · Try Cloud (Free) · Discord · Discussions

This repo ships the client SDKs for the Agent Command Center. The gateway itself — the Go service that handles routing, caching, guardrails, cost tracking, and the OpenAI-compatible endpoint — lives in the future-agi/future-agi platform monorepo. These SDKs give you typed Python and TypeScript clients, plus integrations for LangChain, LlamaIndex, React, and the Vercel AI SDK.

If you already know OpenAI's SDK, you already know these. Swap OpenAI for AgentCC, point base_url at the gateway, and every gateway feature — multi-provider routing, semantic caching, inline guardrails, per-key budgets — is available through the same call.

• OpenAI-compatible surface. Chat, completions, embeddings, images, audio, moderations, files, batches, rerank, responses — same shape as OpenAI's SDK. Migrating is a one-line change.
• 100+ providers through one endpoint. OpenAI, Anthropic, Google, Vertex AI, Bedrock, Azure, Groq, Together, Mistral, Fireworks, Ollama, vLLM — whatever you pick, your SDK call doesn't change.
• 15 routing strategies, surfaced per request. Fallback chains, shadow traffic, latency-aware routing, cost-optimised selection, circuit breakers — configured via a typed config option or gateway-side virtual keys.
• Streaming, tool calling, structured output. Iterator patterns in Python, async iterables in TypeScript — both fully typed.
• Inline guardrails + cost tracking. Every request can carry a guardrail policy and a budget header. The gateway enforces both; the SDK surfaces the results.
• Framework integrations that don't rewrap. @agentcc/langchain is a genuine BaseChatModel. @agentcc/vercel is a real AI SDK provider. @agentcc/llamaindex implements LlamaIndex's LLM and BaseEmbedding. Use them the way you use the originals.

Get a key at app.futureagi.com (free tier available) or self-host the gateway — then:

pip install agentcc

import os
from agentcc import AgentCC

client = AgentCC(
    api_key=os.environ["AGENTCC_API_KEY"],
    base_url="https://gateway.futureagi.com/v1",
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

npm install @agentcc/client

import { AgentCC } from "@agentcc/client";

const client = new AgentCC({
  apiKey: process.env.AGENTCC_API_KEY,
  baseUrl: "https://gateway.futureagi.com/v1",
});

const response = await client.chat.completions.create({
  model: "gpt-4o",
  messages: [{ role: "user", content: "Hello!" }],
});
console.log(response.choices[0].message.content);

_{Set AGENTCC_API_KEY and AGENTCC_BASE_URL in your environment and both clients pick them up automatically.}

Everything the gateway supports — routing strategies, caching, guardrails, budgets — is available per request via a config option. No separate API to learn.

import { AgentCC, type GatewayConfig } from "@agentcc/client";

const config: GatewayConfig = {
  strategy: "fallback",
  targets: [
    { provider: "openai", model: "gpt-4o" },
    { provider: "anthropic", model: "claude-sonnet-4-20250514" },
  ],
};

const client = new AgentCC({
  apiKey: process.env.AGENTCC_API_KEY,
  baseUrl: "https://gateway.futureagi.com/v1",
  config,
});

_{Gateway docs → · Routing strategies → · Guardrails →}

Each integration has its own README with full examples.

These SDKs are one slice of the Future AGI platform — an open-source stack for making AI agents reliable. You can use them standalone against any Agent Command Center deployment, or alongside the rest.

• Python 3.9+ — for agentcc
• Node 18+ — for all @agentcc/* packages
• An Agent Command Center endpoint — either Future AGI Cloud (free tier) or a self-hosted gateway

• Full docs — product overview, concepts, tutorials
• Python SDK reference
• TypeScript SDK reference
• Gateway docs — routing, caching, guardrails, budgets
• Cookbook — end-to-end recipes

Contributions welcome — bug fixes, new framework integrations, examples, docs improvements, anything.

1. Browse good first issue
2. Read the Contributing Guide
3. Say hi on Discord or Discussions

Security reports: see SECURITY.md.

Apache License 2.0 — see LICENSE.