Skip to main content
Fireworks Nexus is the open-weight intelligence layer for agentic coding. It connects the harnesses your teams already use to Fireworks inference, routes each request to the right model, and gives admins per-seat cost controls — without replacing your workflows. You can adopt Nexus as a full platform or use its parts independently. Many teams start with FireConnect or FireRouter on a standard Fireworks API key before rolling out seat-level controls.

​
How Nexus fits together

Fire Pass is not Nexus. Fire Pass is an experimental promotion aimed at individual developers and VIP invites. Nexus is the team control plane for connecting harnesses, intelligent routing, and per-seat spend limits. Both can use FireConnect, but they serve different audiences.

​
Typical paths

1

Connect a harness

Install FireConnect, sign in, and run fireconnect <harness> on. This points your coding tool at Fireworks with sensible model defaults.
2

Add intelligent routing (optional)

Enable FireRouter when you want automatic cost optimization — simple prompts stay on open models, hard prompts pass through to closed-source models you BYOK.
3

Control spend per seat (optional)

For team accounts, enable per-user usage limits to cap serverless spend per developer and block overages until the billing period resets.

​
Use components on their own

Nexus is modular. You do not need every piece to get value:
  • FireConnect only — route harnesses to Fireworks open models with a standard API key (fw_...). No FireRouter or seat limits required.
  • FireRouter only — call the firerouter model directly through the Fireworks inference API, or enable it in FireConnect with --model firerouter.
  • Usage limits only — cap serverless spend per user on supported models without changing how harnesses connect.
FireRouter and FireConnect are available on self-serve (PLG) today. Per-user usage limits require a team account — reach out to your Fireworks contact to enable them.

FireConnect

Install the CLI and route Claude Code, OpenCode, Codex, and more

FireRouter

Automatic routing between open and closed-source models

Usage limits

Per-user spending caps for serverless inference

Fire Pass

Invite-only promotion for personal agentic coding

​
Learn more