The production layer for AI agents.

Build, deploy, observe, and govern multi-agent AI systems — in any framework, on any cloud. Open-core. Production-grade. No lock-in.

bash
npm install @circuitnova/core

Trusted by engineers building at

Acme
Linear
Vercel
Notion
Stripe
Supabase
The problem

Building an agent demo takes a weekend.
Shipping it to production takes six months.

Framework lock-in

AutoGen migrating to Microsoft Agent Framework. LangGraph 1.0 just shipped. Every choice is a future migration.

Observability black hole

When your agent fails in production, debugging is forensic work — replay missing, traces vague, costs hidden.

Cost opacity

Five-agent workflows can burn through your LLM budget overnight, and you won't see it until the invoice arrives.

The solution

One runtime. Every framework. Full visibility.

circuitnova ai is the connective tissue between your agent code and production reality.

circuitnova • observability

Recent runs

  • run_8c2aOK
  • run_8c29OK
  • run_8c28RETRY
  • run_8c27OK
  • run_8c26FAIL
  • run_8c25OK

research-agent / run_8c2a

4.21s · $0.0182 · 12 steps

OK
planner
1.10s$0.0042
search.tavily
0.85s$0.0009
search.arxiv
1.02s$0.0011
writer
2.26s$0.0131
critic
0.45s$0.0021

Framework-agnostic runtime

Run LangGraph, CrewAI, AutoGen or our SDK on a single runtime.

Native observability

Per-step traces, replays, and diffs out of the box.

Durable execution

Long-running agents survive crashes, retries, and restarts.

Cost intelligence

Token, latency, and dollar attribution per run, per agent, per tool.

Open standards

MCP and A2A native — no proprietary protocol tax.

How it works

From prototype to production. No rewrites.

Three steps. The same code path from your laptop to the edge.

01

Build

Use LangGraph, CrewAI, AutoGen, or our native SDK. Pick the framework you love — we run all of them.

ts
import { agent } from "@circuitnova/core";

export const researcher = agent({
  framework: "langgraph",
  graph: myGraph,
});
02

Run

Deploy to our cloud, your cloud, or self-host. One command. Durable execution included.

shell
$ circuitnova deploy researcher

✓ uploaded  •  ✓ provisioned  •  ✓ live
→ https://api.circuitnova.ai/r/researcher
03

Observe

Every run is traceable, replayable, and debuggable. Down to the tool call, token, and dollar.

trace
run_8c2a • OK • 4.21s • $0.0182
├─ planner          1.10s   $0.0042
├─ search.tavily    0.85s   $0.0009
└─ writer           2.26s   $0.0131
Integrations

Framework-agnostic by architecture, not by marketing.

circuitnova ai is built on adapter-pattern primitives. Run any framework, on any cloud, with the same runtime, the same observability, the same deployment.

LangGraph
CrewAI
AutoGen
OpenAI
Anthropic
Google
AWS
Azure
GCP
$14B+
AI orchestration market by 2026
22% CAGR
Projected growth through 2034
60%
Fortune 500 using agent frameworks
MIT
License — open source forever
Pricing

Start free. Scale when you need to.

Open-source core. Hosted plans when production calls.

Core

Freeforever

Self-host the open-source runtime.

  • MIT license
  • Single-node runtime
  • Local traces
  • Community support
Most popular

Pro

$49/user/mo

For shipping engineers.

  • 250k runs / mo
  • 30-day traces
  • Cost dashboards
  • Slack + email support

Business

$1,499/mo

For teams in production.

  • 25 seats
  • 10M runs / mo
  • SSO + RBAC
  • 99.9% SLA

Ready to ship agents that actually work?

Join thousands of developers building production AI on circuitnova ai.