Olywork gives an agent one authenticated connection to 3,700+ catalog tools across 126 providers. Credentials are injected server-side, and plain proxy calls relay the upstream response without transforming it.

Each catalog endpoint shows its price before execution. Charges use integer micro-USD, failed calls are not billed, and new teams start with $1.00 of credit. Your own registered tools are not metered by Olywork.

No. Your single Olywork token authenticates your agent across every provider in the catalog. If you already have your own key for a provider, register it once with olywork keys set <provider> <key> — your key always wins and those calls are never metered through us.

Yes. Olywork handles OAuth flows natively for supported providers. Your users authorize once, and your agent is immediately able to act on their behalf. The raw token is never exposed to your code — Olywork injects it server-side and relays responses verbatim.

olywork install registers the MCP server into Claude Code, Cursor, and opencode automatically. Any system that can make an HTTP request — LangChain, AutoGen, CrewAI, raw curl — can use the REST endpoint directly.

Olywork does not add rate limits beyond the upstream provider's own. If the provider throttles, the call returns the provider's error verbatim — and that call is not billed. You can set daily spend caps per tag or per team from your dashboard.

Yes. Paste any OpenAPI 3.x or Swagger 2.0 specification URL into your dashboard or run olywork tool import-openapi <url>. Olywork automatically parses all endpoints and generates authenticated tools and usage recipes for your agent.

Yes. Olywork can be deployed as a self-hosted Docker container or Python server inside your private VPC with local SQLite or PostgreSQL storage, ensuring total network isolation and zero key leakage.