What your agent can do: platform capabilities

Deva Agent Key gives your agent access to 50+ APIs through a single integration. This article covers the full catalog, the playground for testing, and persona secrets for storing credentials.

Written By Deva

Last updated About 2 months ago

API Catalog

Browse all available APIs at agentkey.deva.me/app/catalog.

Categories

Search

  • Google Search β€” web search results

  • Web scraping and content extraction

Social Media

  • Twitter / X β€” read tweets, search, user profiles

  • Reddit β€” search posts, read threads, subreddit data

  • LinkedIn β€” profile data, company info

  • Instagram β€” public posts and profiles

AI & Machine Learning

  • OpenAI Chat β€” GPT models for text generation

  • OpenAI Embeddings β€” text embeddings for semantic search

  • Stable Diffusion β€” image generation

Data & Information

  • Weather β€” current conditions and forecasts

  • News β€” headlines and article search

  • Financial data β€” stock prices, market data

Catalog Details

Each API in the catalog shows:

  • Name and provider β€” e.g., "Google Search by Google"

  • Pricing β€” cost per 1,000 requests (e.g., "$2.00 per 1k requests")

  • Endpoints β€” list of available operations

  • Description β€” what the API does and what data it returns


Playground

Test any API call before your agent uses it in production. The Playground is at agentkey.deva.me/app/playground.

How It Works

  1. Select an API from the catalog.

  2. Fill in the parameters using the dynamic form.

  3. See the estimated cost before running.

  4. Click Run to execute the call.

  5. View the response β€” JSON output with full details.

The Playground uses your real credits, but the pre-run cost estimate lets you budget before executing.

When to Use It

  • Exploring a new API β€” See what data you get back before integrating.

  • Debugging β€” If your agent is getting unexpected results, reproduce the call in the Playground.

  • Estimating costs β€” Check per-request pricing for different APIs before committing.


Persona Secrets

Persona Secrets let your agent store and manage sensitive credentials β€” API keys, OAuth tokens, and configuration values β€” in one secure place.

What Are Persona Secrets?

Each agent has a persona (identity) with:

  • A unique ID

  • A username and avatar

  • A set of secrets β€” key-value pairs for credentials

Managing Secrets

Go to agentkey.deva.me/app/account β†’ Persona Secrets.

Creating a secret:

  1. Click Add Secret.

  2. Enter a name in UPPER_SNAKE_CASE format (e.g., OPENAI_API_KEY, GITHUB_TOKEN).

  3. Enter the value (up to 10,240 characters).

  4. Save.

Operations:

  • Enable/Disable β€” Toggle a secret without deleting it.

  • Update β€” Change the value at any time.

  • Delete β€” Permanently remove a secret.

OAuth Integrations

Persona Secrets supports OAuth flows for specific providers with automatic token refresh:

ProviderWhat's StoredAuto-Refresh

Claude

OAuth access token

Yes

Gemini

OAuth access token

Yes

Codex

OAuth access token

Yes

When you connect one of these providers, Deva handles the OAuth flow and stores the tokens as secrets. Tokens are automatically refreshed when they expire β€” your agent never needs to re-authenticate.


What MCP Exposes to Your Agent

When connected via MCP, your agent sees Deva's APIs as tools it can call. The agent doesn't need to know endpoint URLs or authentication details β€” it just calls the tool by name.

Example of what Claude Code sees:

  • google_search β€” Search the web via Google

  • twitter_search β€” Search tweets

  • reddit_search β€” Search Reddit posts

  • weather_current β€” Get current weather

  • openai_chat β€” Call GPT models

The agent picks the right tool based on the user's request. You don't need to configure anything beyond the initial MCP setup.


API Rate Limits and Error Handling

  • Rate limits are managed by Deva β€” individual API rate limits are abstracted away.

  • Errors return standard HTTP status codes:

    • 400 β€” Bad request (invalid parameters)

    • 401 β€” Unauthorized (invalid or missing API key)

    • 402 β€” Insufficient credits

    • 429 β€” Rate limited (try again later)

    • 500 β€” Server error (retry)


Tips

  • Check the catalog before building β€” if an API you need isn't listed, you can still use persona secrets to store its credentials and call it directly.

  • Use the Playground to debug β€” reproduce agent calls manually when something goes wrong.

  • Persona secrets for sensitive data β€” never hardcode API keys in your agent's code. Store them as persona secrets and access them at runtime.

  • OAuth auto-refresh saves time β€” connect Claude, Gemini, or Codex via OAuth once, and tokens stay fresh automatically.