Introducing the Deva Agent API

One API for agentsmodels, tools, and credits

Give your agent one Deva key for frontier models, web search, images, voice, browser tools, and usage metering across providers.

Deva Agent API

Built for any agent, anywhere

Agent teams do not need another API directory. They need one runtime surface for model calls, tools, metering, and spend control.

Agent resources, one subscription

  • 34 frontier models through deva.me/models
  • Web search, image generation, voice, and browser tools
  • Karma metering, monthly credits, and top-ups
  • Hosted Genie when you want Deva to run the agent too

Bring your own agent.

Point OpenClaw, Hermes, or your own runtime at one Deva endpoint.

One API key.

Models, tools, credits, and limits sit behind the same credential.

One bill.

Use Deva credits across providers instead of managing separate vendor spend.

No hard lock-in.

Use Deva as the default path, then bring provider keys when you need them.

Agent surfaces

One API key where your agents operate

01

Get one Deva key

Create a Deva account and use the same credential for model calls, tools, and metering.

02

Point your agent at Deva

Use Deva from OpenClaw, Hermes, or your own runtime without juggling provider-specific API keys.

03

Add Genie only if you want hosting

Run the agent yourself, or let hosted Genie provision the server and connect channels for you.

Deva logo
Slack
Telegram
WhatsApp

Connect agent surfaces

Use one API path for channels like Slack, Telegram, and WhatsApp when your agent needs to respond.

Deva

Route across models

Call Claude, OpenAI, Kimi, and more through one Deva credential instead of provider-specific wiring.

Hey, I need help scheduling a team meeting that works well for everyone. Any suggestions for finding an optimal time slot?

User Avatar
Deva logo

Meter the whole run

Track credits across model calls and tools so agent spend is visible from the first request.

Agent API pricing

Pick a plan for your key. It works with any agent you run.

Plus

Monthly credits for every model and tool

$20/ month
Get your key
  • 34 frontier model catalog
  • Search, images, voice, and browser tools
  • Works with OpenClaw, Hermes, or any agent
  • Free to try: 5k bronze karma daily
Most Popular

Super

Larger credit pool for daily drivers

$100/ month
Get your key
  • Everything in Plus
  • 5x the monthly credits
  • Top up anytime with karma
  • Free to try: 5k bronze karma daily

Ultra

Highest credit pool for heavy agents

$200/ month
Get your key
  • Everything in Super
  • Highest monthly credits
  • Top up anytime with karma
  • Free to try: 5k bronze karma daily

Do not want to self-host? Genie runs the agent for you.

Your Deva key pre-wired on a hosted server, live in under a minute.

Start hosted

Genie Go

2GB / 2vCPU / 60GB

$20/mo

Genie Plus

4GB / 2vCPU / 80GB

$40/mo

Genie Pro

8GB / 2vCPU / 160GB

$65/mo

Comparison

Multi-provider wiring vs Deva

Traditional

  • Creating provider accounts15 min
  • Issuing API keys10 min
  • Adding search and media APIs10 min
  • Wiring browser tools10 min
  • Normalizing usage limits10 min
  • Reconciling billing5 min
Total60 min
Deva

1 API

Give your agent one Deva key. We route model calls, agent tools, credits, and hosted Genie when you want the server too.

Feature

API catalogs

Directories like getanyapi.com or monid.ai help discovery, but your agent still needs glue across providers.

Deva Agent API

One API for agents: models, tools, credits, and a hosted path when you want Deva to run it.

Single agent API surface

Frontier models plus agent tools

Directory

One account for metering and top-ups

Works with self-hosted agents

Hosted Genie on-ramp

Bring your own provider keys

Built for production agent spend

Privacy & data

Built so your data stays yours

How Deva approaches credentials, routing, and control so you can ship agent API traffic with confidence.

01

Scoped API credentials

Use a dedicated Deva credential for agent workloads and rotate access when your policy requires it.

02

Encryption by default

Traffic is protected in transit with modern TLS. Stored data uses platform-standard encryption at rest.

03

No training on your prompts

Your agent request content is not used to train public foundation models. It stays in your environment.

04

You control tools

Enable only the model and tool surfaces you need. Revoke or rotate access anytime from your account.

05

Bring your own keys

When you use your own API keys, they are stored and applied securely without exposing them in the client.

06

Usage visibility

Track credits and top-ups from one Deva account instead of reconciling separate provider bills.

07

Least-privilege access

Operational access to infrastructure follows strict controls and is limited to what is required to run the service.

08

Transparent boundaries

We document what runs where, what leaves your instance, and how subprocessors are used so you can review risk.

FAQ

Frequently Asked Questions

NewGet started today

Start with one API for agents

Connect your runtime to Deva for models, tools, credits, and usage metering. Add hosted Genie only when you want us to run it.