Skip to main content
For Developers

One SDK. Every AI model.

Betwixt is an AI agent platform. Bring your own API keys from Claude, GPT, Gemini, or any provider — Niko orchestrates them behind one SDK with persistent memory and up to 96% fewer tokens.

app.ts
 
How It Works

You write one API call. Betwixt does the rest.

Niko, your AI orchestrator, coordinates Claude, GPT, Gemini, and Codex behind the scenes. Memory persists. Token costs drop. Your code stays simple.

Agent RunsLive
succeededClaudefeat/auth-refactora3f2c912s
succeededCodexfix/db-migrationb7e1d434s
runningGeminiui/dashboard-v2c9a8f2...
succeededNikomaind1f3e58s
failedClaudefeat/memory-apie2a4b641s
succeededCodexchore/deps-updatef5c7d819s
Your Agents
NikoOrchestratorOnline
ClaudeCode & StrategyReady
GeminiFrontendReady
CodexBackend & InfraReady
// With Betwixt — one call, auto-routed, with memory
import { betwixt } from '@betwixt/sdk';

const result = await betwixt.chat({
  model: 'auto',     // auto-picks the best model
  memory: true,       // remembers past context
  messages: [{ role: 'user', content: prompt }],
});
5Managed APIs
4+AI Providers
<50msRouting Latency
100%OpenAI Compatible
APIs

OpenAI-compatible. Drop in and go.

Replace your OpenAI base URL with Betwixt. Get memory, smart routing, and multi-model access — same SDK, same types.

POST /v1/memory/storeStore and retrieve agent context with semantic search
Request
curl https://api.betwixt.dev/v1/memory/store \
  -H "Authorization: Bearer bx_key" \
  -d '{"content": "Auth uses RS256 JWT", "type": "decision"}'
Response
{
  "id": "mem_a8f3c2",
  "status": "stored",
  "embedding_dims": 384,
  "searchable": true
}
200 OKPOST /v1/memory/store23ms
{ "id": "a3f2c9", "status": "success" }
200 OKPOST /v1/llm/chat/completions341ms
{ "id": "b7e1d4", "status": "success" }
200 OKPOST /v1/workflows/runs17ms
{ "id": "c9a8f2", "status": "success" }

Every call returns structured JSON. Every response is logged.

Platform

The full stack for AI-native apps.

Git Repos & CI/CD

Full git hosting with branches, PRs, CI pipelines, webhooks, and Hive publishing. Clone via HTTPS or SSH.

Model Hub

Upload, version, and deploy AI models. PyTorch, TensorFlow, ONNX, GGUF, Safetensors. Inference API and code snippets.

Datasets

Manage training data. CSV, JSONL, Parquet, Arrow with schema detection and tabular preview.

Agent Orchestration

Manage Niko, Claude, Gemini, Codex. Real-time activity feeds, status monitoring, task routing.

Infrastructure

Connect AWS, GCP, Azure, Hetzner. Manage VPS servers, domains, DNS, CDN, and SSL from one place.

Scheduled Agents

Automate agent runs on a schedule. Set triggers, define workflows, and let agents work autonomously.

Two Modes

Simple when you want it. Powerful when you need it.

SimpleDefault

Your apps and automations.

  • Running applications
  • Automation status
  • Talk to Niko
  • Browse the Hive
DeveloperToggle on

The full engineering workspace.

  • Everything in Simple, plus:
  • Git repos, PRs, CI/CD
  • Model hub & datasets
  • 5 managed APIs with scoped keys
  • Infrastructure & cloud providers
  • Scheduled & triggered automations

Install bx. Bring your keys. Ship.

Bring your own API keys from any provider — Claude, GPT, Gemini — and start building in under a minute.

Stop wasting tokens on context your agents already learned.

Get Started
curl -fsSL https://get.betwixtai.com | sh