Nexus
Talk to us

Own your AI factory.One gateway, every model.

Every person and coding agent on every model, through your keys and your controls.

Talk to us
Any model
ClaudeChatGPTGeminiGrokDeepSeekLlamaMistralQwenComing soon
Works with
Claude CodeCodexGitHub
Backed by
LightspeedPanteraDragonflyFaction
Team from
Stanford UniversityBrown University
Tree leaf shadows cast on a smooth light grey stone wall
AccountSpendSample
Spend
Your budget, the spend limit of each person, and what every model call behind it cost.
Organization budget$4,000.00This month
Spent$1,412.60
Reserved$3.84Held for calls in flight
Remaining$2,583.56Resets on the 1st
Spend limits this monthDefault $50.00 per person · 1 of 14 at their limit
PersonTop modelSpent of limitStatus
Backend engineerclaude-sonnet-5.5
$38.20 / $50
Within limit
Data engineergpt-6.1
$64.10 / $120
Override
Support leadclaude-haiku-5.5
$30.00 / $30
At limit
Product designerclaude-opus-5.5
$15.40 / $50
Within limit

One gateway, yours.

Your people and their coding agents call every model through one gateway, on your own provider keys. Policies, budgets and routing apply to every call, and each one is attributed to a person, a team and a model.

Your keysyour organization, your controls
People and agents
Nexus UIChat, spend, models and teams
The nexus CLInexus claudenexus codex
The APIOne endpoint for every model
MCPFor the agents you already run
Nexus GatewayEvery call checked, routed and recorded before it reaches a provider.
Model policiesallow, block
Budgets and spend limitsrefused before the call
Routingyour keys
Audit logappend-only
Model providers
AnthropicClaude models
OpenAIGPT and Codex models
Coming soonGoogle, xAI, DeepSeek, Meta, Mistral, Qwen

Claude Code and Codex, through your gateway.

Type nexus claude instead of claude. Nothing else changes for your engineers, and every call runs on your keys, within your policies and budgets.

~/acme/api nexus claude
~/acme/api $ nexus claude --team platform
Nexus Gateway · session sess_4kQ9xTmR2bWz · claude-opus-5.5 · team platform · every call reserved, metered and checked
> Add retries with backoff to the webhook sender, then run its tests
● Read(src/webhooks/sender.ts) └ Read 84 lines
● Update(src/webhooks/sender.ts) └ 18 additions, 3 removals
● Bash(npm test -- webhooks) └ 12 passed
● Retries now back off exponentially, three attempts at most, and the tests pass.
✓ claude exited 0 after 6 min 12 s · spend: $0.84 Sample → nexus-factory-web.vercel.app/app/acme/sessions/sess_4kQ9xTmR2bWz

Know what AI costs you, person by person.

Every call through the gateway is attributed to a person, a team and a model, so the bill stops being one number nobody can explain.

1Add your provider keys
2Point your people and agents at the gateway
3See spend by person, team and model
SpendAcme · October, so far
14 people · 3 teamsSample
Spent across every model
$1,412/ $4,000
$0Organization budget$4,000
PersonModelTeamSpent
Backend engineerclaude-opus-5.5Platform$312.40
Data engineergpt-6.1Data$264.10
Mobile engineerclaude-sonnet-5.5Platform$198.75
Support leadclaude-haiku-5.5Operations$30.00
Generated dither texture
Own your intelligence. Own your factory.

It runs on your keys, under your controls. Export everything and stop at any time.

Your keys, your models

Bring your own keys for every provider. Your models and vendor contracts stay yours.

Every call attributed

Each call is recorded with the person, team and model behind it, and what it cost.

Your data stays yours

Chats, artifacts and results stay inside your organization, under its permissions and audit log.

Export everything

Every config and artifact is yours to export, and you can stop at any time.

Start with an API key.

Talk to us and our engineers will set it up with you: add your provider keys, point your first agent at the gateway, and see every call it makes.

Talk to us
API keysCreate key
platform-ciJust now
nx_live_••••••••••••7Q2c
Team platform · models your policy allows
backend-laptopLast used 2 h ago
data-notebooksLast used yesterday