oauth handled automatically

Integrating with third-party services is painful. You need to implement OAuth flows, handle token refresh, manage secure storage, and deal with rate limits. For every single service.

inference.sh handles all of this for you. When your agent needs to access Google Drive, Slack, or Salesforce, we manage the entire OAuth flow. Tokens are stored securely, refreshed automatically, and your agent just works.

Your users authenticate once, and your agents can access their connected services forever. No integration code required.

150+ pre-built apps

We've built integrations with the services your agents actually need:

• productivity. Google Workspace, Microsoft 365, Notion, Slack, Discord
• crm & sales. Salesforce, HubSpot, Pipedrive
• communication. WhatsApp, Telegram, Twilio, SendGrid
• development. GitHub, GitLab, Jira, Linear
• data. Airtable, PostgreSQL, MongoDB, Snowflake
• finance. Stripe, QuickBooks, Plaid

Each integration is a tool your agent can use. No API wrappers to write, no authentication to handle, no edge cases to debug.

browse all apps →

bring your own keys

Already have credits with cloud providers? Use them. BYOK lets you connect your existing accounts:

• google cloud. Use your GCP credits for Gemini and Vertex AI
• microsoft azure. Use Azure OpenAI Service with your existing subscription
• amazon web services. Use Bedrock with your AWS billing
• direct api keys. OpenAI, Anthropic, Mistral, Deepseek—use your own keys

why byok matters

Enterprise cloud agreements often include significant AI credits. Your company might already have $100k in Azure credits sitting unused. BYOK lets you put those credits to work.

It also means full control over billing. See exactly what you're spending on each provider. Optimize costs by routing different workloads to different providers based on their pricing.

And for compliance, some organizations require that AI calls go through their own cloud accounts for audit and governance reasons. BYOK makes this possible without giving up the agent runtime.

how it works

Connect your credentials once in the dashboard. When your agent makes a request:

1. We route the request through your connected provider
2. The provider bills your account directly
3. You get full observability in inference.sh
4. Your agent continues as normal

Switch between BYOK and our managed inference at any time. Mix and match based on cost, latency, or availability.

custom integrations

Need to connect to an internal service or a provider we don't support yet? Build custom tools that your agents can use. Full TypeScript/Python SDK support with the same OAuth and credential management infrastructure.

Or reach out—we're constantly adding new integrations based on customer needs.

get started

Browse available integrations in the app store. Connect your OAuth credentials or API keys in the dashboard. Your agents can start using them immediately.

explore integrations →