Agent skills for integrating You.com's AI-powered search, content extraction, and web capabilities with popular AI agent frameworks and bash-based agents.
These skills provide interactive workflows that guide your AI agent through setting up You.com integrations for SDKs, frameworks, and CLI tools.
Integrate Vercel AI SDK applications with You.com tools for real-time web search, AI-powered answers, and content extraction.
Use when:
- Building AI SDK applications with
generateText()orstreamText() - Adding web search capabilities to your AI agents
- Extracting and processing web content programmatically
Features:
- Interactive setup workflow for existing or new projects
- Three powerful tools:
youSearch,youContents - Smart integration with existing AI SDK code
- Support for multiple AI providers (Anthropic, OpenAI, Google, etc.)
Connect Claude Agent SDK (Python and TypeScript) to You.com's HTTP MCP server for web search and content extraction.
Use when:
- Building Claude-powered agents in Python or TypeScript
- Integrating MCP tools with Claude Agent SDK v1 or v2
- Adding You.com capabilities to existing Claude applications
Features:
- Complete templates for Python and TypeScript (v1 & v2)
- HTTP MCP server configuration patterns
- Bearer token authentication setup
- Error handling and validation examples
Add You.com MCP tools to OpenAI Agents SDK using Hosted MCP or Streamable HTTP modes.
Use when:
- Building OpenAI-powered agents with MCP integration
- Using Python or TypeScript OpenAI Agents SDK
- Choosing between Hosted MCP and Streamable HTTP deployment
Features:
- Dual-mode templates (Hosted MCP + Streamable HTTP)
- Python and TypeScript implementations
- Complete configuration examples for both modes
- Tool approval and validation patterns
Integrate You.com's remote MCP server with crewAI agents for web search, AI-powered answers, and content extraction.
Use when:
- Building crewAI agents that need real-time web access
- Integrating You.com MCP via
MCPServerHTTPorMCPServerAdapter - Adding web search and content extraction to existing crewAI workflows
Features:
- DSL and MCPServerAdapter integration patterns
- Python implementation with uv/pip setup
- Bearer token authentication for the remote MCP server
- Complete crewAI crew and task configuration examples
Integrate LangChain applications (TypeScript and Python) with You.com tools for web search, content extraction, and retrieval.
Use when:
- Building LangChain.js agents with
createAgentandinitChatModel(TypeScript) - Using
YouRetriever,YouSearchTool, orYouContentsToolwith LangChain (Python) - Adding web search or content extraction to existing LangChain workflows
Features:
- TypeScript:
youSearchandyouContentsvia@youdotcom-oss/langchain, structured output with Zod - Python:
YouRetrieverfor RAG chains,YouSearchTool+YouContentsToolfor agents vialangchain-youdotcom - Prompt injection defense guidance (W011 trust boundary)
- Direct invocation and agent-based usage patterns for both languages
Use @youdotcom-oss/teams-anthropic to add Anthropic Claude models (Opus, Sonnet, Haiku) to Microsoft Teams.ai applications. Optionally integrate You.com MCP server for web search and content extraction.
Use when:
- Building Teams.ai apps with Claude models
- Need streaming, function calling, or conversation memory
- Optionally want web search capabilities via You.com MCP
Features:
- Two paths: Basic setup (Claude only) or with You.com MCP
- Complete templates for new and existing apps
- Streaming responses and function calling
- Conversation memory with Teams.ai Memory API
Integrate You.com APIs (Research, Search, Contents) into any language using direct HTTP calls — no SDK required.
Use when:
- Calling You.com APIs directly without an SDK wrapper
- Need synthesized, cited answers via the Research API
- Building custom search pipelines with raw Search + Contents data
- Working in a language without a dedicated You.com SDK
Features:
- Research API: one call for multi-step reasoning with cited Markdown answers
- Search API: raw web and news results with filtering, pagination, and livecrawl
- Contents API: full page extraction (HTML, Markdown, metadata) from any URL
- Language-agnostic — works with any HTTP client (fetch, requests, httpx, curl)
- TypeScript and Python reference implementations included
- JSON Schemas for all request/response shapes
Web search, research with citations, and content extraction for bash agents using curl and You.com's REST API.
Use when:
- Working with bash-capable AI agents (Claude Code, Cursor, Codex, etc.)
- Need fast web search with verifiable citations
- Want simultaneous search + content extraction (livecrawl)
- Building bash agent workflows with curl and jq
Features:
- Search works without an API key (free tier)
- Livecrawl: search + extract content in one API call
- Research with citations at multiple effort levels
- Compatible with any bash-based agent
Install All Skills (recommended):
# Using npm
npx skills add youdotcom-oss/agent-skills
# Using Bun (recommended)
bunx skills add youdotcom-oss/agent-skillsThis installs all 8 skills at once:
ydc-ai-sdk-integrationydc-claude-agent-sdk-integrationydc-openai-agent-sdk-integrationydc-crewai-mcp-integrationydc-langchain-integrationteams-anthropic-integrationyoudotcom-apiyoudotcom-cli
Install Individual Skills:
# Install just one skill
npx skills add youdotcom-oss/agent-skills --skill youdotcom-cli
bunx skills add youdotcom-oss/agent-skills --skill ydc-ai-sdk-integration
# Install multiple specific skills
npx skills add youdotcom-oss/agent-skills --skill youdotcom-cli --skill ydc-ai-sdk-integrationBefore using any skill, you'll need a You.com API key:
- Get API Key: Visit you.com/platform/api-keys
- Set Environment Variable:
export YDC_API_KEY="your-api-key-here"
- Request Integration: Ask your AI agent to integrate You.com (see Usage examples below)
Once installed, your AI coding agent will automatically activate the relevant skill when you request integration. For example:
- "Integrate Vercel AI SDK with You.com tools"
- "Set up Claude Agent SDK with You.com MCP"
- "Add You.com to my Teams app with Anthropic"
- "Configure OpenAI Agents SDK with You.com MCP"
- "Integrate You.com MCP with my crewAI agents"
- "Add You.com tools to my LangChain.js agent"
- "Integrate You.com Research API into my Python app"
- "Call You.com Search and Contents APIs directly with fetch"
- "Add You.com CLI tools to my bash agent"
Each skill provides step-by-step instructions, code templates, and validation checklists.
Each skill follows the agent-skills-spec format:
skills/{skill-name}/
├── SKILL.md # Complete workflow with YAML frontmatter
└── assets/ # Code templates (optional, mostly inlined)
Skills are self-contained:
- YAML frontmatter defines skill metadata (name, description, category, keywords, compatibility)
- Markdown body contains complete workflow, inline code examples, templates, validation, and troubleshooting
- Assets directory (optional) for additional templates - most examples are now inlined for immediate visibility
API Keys:
- You.com API key: Get yours
- Provider API keys (Anthropic, OpenAI, etc.) depending on the skill
Each skill includes:
- Prerequisites - Required packages and environment variables
- Complete templates - Ready-to-run code for Python and TypeScript
- Configuration examples - Side-by-side comparisons for different modes
- Validation checklists - Ensure your integration works correctly
- Troubleshooting - Common issues and solutions
Create a .env file in the project root with the following API keys:
# Required for all skills
YDC_API_KEY=your-you-com-api-key
# Required for Claude Agent SDK skill
ANTHROPIC_API_KEY=your-anthropic-api-key
# Required for OpenAI Agent SDK skill
OPENAI_API_KEY=your-openai-api-keyGet API keys from:
- You.com: you.com/platform/api-keys
- Anthropic: console.anthropic.com
- OpenAI: platform.openai.com/api-keys
Skills are validated by running Claude Code against prompts and checking that the generated integration code passes real API tests.
Run all skill evals:
bun run evalRun a single skill eval:
bun run eval --skill ydc-crewai-mcp-integrationRun with parallelism:
bun run eval -j 4Regenerate data/RESULTS.md from existing results (no re-run):
bun run eval:summaryNote: Evals use real API keys from .env and invoke Claude Code as a subprocess to generate integration code. Valid API keys are required.
data/
├── prompts/
│ └── prompts.jsonl # One entry per skill variant (id, prompt, grader config)
├── results/
│ └── results.jsonl # Grader output per eval run (gitignored)
└── RESULTS.md # Human-readable summary (committed on weekly CI run)
tests/{skill-id}/ # Generated integration code lives here (gitignored)
├── agent.ts # Example: TypeScript integration file
└── agent.spec.ts # Tests that validate the generated code
scripts/
├── run.ts # Eval orchestrator (clean → harness → grade → summarize)
└── grader.ts # Scoring logic for generated integration code
Eval IDs and test directories use language suffixes where needed:
ydc-claude-agent-sdk-integration-python→tests/ydc-claude-agent-sdk-integration-python/ydc-claude-agent-sdk-integration-typescript→tests/ydc-claude-agent-sdk-integration-typescript/ydc-openai-agent-sdk-integration-python→tests/ydc-openai-agent-sdk-integration-python/ydc-openai-agent-sdk-integration-typescript→tests/ydc-openai-agent-sdk-integration-typescript/youdotcom-api-python→tests/youdotcom-api-python/youdotcom-api-typescript→tests/youdotcom-api-typescript/youdotcom-cli→tests/youdotcom-cli/- Single-variant skills (e.g.,
ydc-crewai-mcp-integration) use a single test directory
Workflow:
data/prompts/prompts.jsonlcontains prompts that trigger each skill- The eval harness runs Claude Code against each prompt, generating code into
tests/{skill-id}/ - The grader validates the generated code against the test files
- Results are written to
data/results/results.jsonland summarized indata/RESULTS.md
Evals run automatically on:
- Pull requests that change
skills/*/SKILL.md, assets, or eval scripts - Pushes to main for the same paths
- Weekly schedule (Monday 06:00 UTC) — results committed back to
data/RESULTS.md
Current eval results: see data/RESULTS.md
Check for issues:
# Check all files (TypeScript + Python)
bun run check
# Check only Python files
bun run check:pyAuto-fix issues:
# Fix all files (TypeScript + Python)
bun run check:write
# Fix only Python files
bun run check:write-pyTools:
- TypeScript: Biome for linting and formatting
- Python: Ruff for linting and formatting
- Python package management: uv
- Bun >= 1.2.21 (for TypeScript evals and orchestration)
- Python >= 3.12 (for Python skill evals)
- uv (automatically used by Bun scripts for Python)
Contributions are welcome! To add a new skill:
- Fork this repository
- Create a new skill directory in
skills/ - Add
SKILL.mdfollowing agent-skills-spec format - Add optional assets in
assets/subdirectory - Add a prompt entry to
data/prompts/prompts.jsonland reference test files intests/ - Test your skill with
npx skills add <your-fork> - Submit a pull request
Skill naming convention:
- Directory name must match
namefield in YAML frontmatter - Use kebab-case (e.g.,
ydc-ai-sdk-integration)
MIT - See LICENSE file for details
- Issues: GitHub Issues
- Email: support@you.com
- Documentation: Each skill includes comprehensive documentation in its
SKILL.mdfile