AgentWorks
GitHub Book a call

Product

Overview Goals Crews Relays Brain Vault LLM Gateway Code Auto-improve Connectors Bullpen

Use cases

All premade agents E-commerce operations Sales & GTM Shopify stores Customer support Finance & billing Marketing & growth Engineering ops

Enterprise

Overview Release quality Incident response Security testing Cloud cost Growth analytics Code Expert Crews Vault Pricing Docs GitHub Book a call

LLM Gateway Enterprise Coming soon

One endpoint for every model. Every call checked and costed.

Give every person and internal app one way to call AI models. You choose who can use which model, and see the cost of every call by person and by model.

Book a call See use cases

ModelsIllustrative

ClaudeEngineering, Support
Allowed
GPTEveryone
Allowed
GeminiData team only
Limited

Cost this weekIllustrative

R
RioClaude · support tool
$42
S
SageGemini · data notebooks
$31
O
Docs botGPT · internal app
$12
A
AvaClaude · coding
$27

Use cases

One door for model calls. Illustrative examples.

  • Internal app

    Support tool

    1. Calls Claude with Rio's token
    2. Support may use Claude

    Answer returned, cost recorded

  • Model not allowed

    An intern's script

    1. Asks for the largest model
    2. Not in their allowed models

    Refused and recorded

  • Lost laptop

    An old token

    1. Tries to list the models
    2. Token revoked

    Refused

  • No provider keys in apps

    Any app

    1. Uses the company endpoint
    2. Signs in with a personal token

    Provider keys stay off laptops

  • Monthly cost review

    Finance

    1. Opens the cost summary
    2. By person and by model

    AI spend explained, line by line

  • With Vault

    Your server

    1. Vault governs the tools
    2. LLM Gateway governs the models

    One place for AI access

How it works

Three steps.

  1. Add your models

    Connect the model providers your company uses, once.

  2. Choose who uses what

    Allow each person the models they need.

  3. Hand out tokens

    People and apps call one endpoint. Every call is checked and costed.

A token per personModel access per personEvery call recorded with its costRuns on your server

FAQ

Questions, answered.

Is it available?

Not yet. LLM Gateway is coming soon. Book a call and we will set it up with you when it opens.

How do apps call it?

Through one endpoint on your AgentWorks server: list the models, then send messages or responses requests. Each person or app uses their own token.

What can a token do?

Only list models and run them. It can't change settings or reach the admin screens, and the person's current model access still applies on every call.

Does it replace Vault?

No. Vault governs MCP tools and shared secrets. LLM Gateway governs model calls. They run side by side on your server.

Bring model calls under one roof. Before the bills surprise you.

Tell us which models and apps your teams use, and we will set up LLM Gateway with you.

Book a call See Vault