# Create Your First Agent Step-by-step guide to creating an AI agent on TeamDay — from a simple assistant to a fully equipped team member with skills, MCPs, and subagents. # Create Your First Agent This guide walks you through creating an AI agent on TeamDay — from a quick assistant to a fully equipped team member with skills, MCP integrations, and subagents. Agents are AI employees. They have a name, role, system prompt, model, and equipment. You hire them, equip them, and put them to work. --- ## Quick Start The fastest way to get an agent running. ### Via Web App 1. From the org home page, click **+ Create Agent** 2. Fill in: - **Name**: "Research Assistant" - **Role**: "Research and summarization" - **System Prompt**: "You are a research assistant. Find information, summarize it clearly, and cite your sources." 3. Start chatting ### Via CLI ```bash teamday agents create \ --name "Research Assistant" \ --role "Research and summarization" \ --system-prompt "You are a research assistant. Find information, summarize it clearly, and cite your sources." ``` ### Via API ```bash curl -X POST "https://cc.teamday.ai/api/v1/agents" \ -H "Authorization: Bearer $TEAMDAY_API_TOKEN" \ -H "Content-Type: application/json" \ -d '{ "name": "Research Assistant", "systemPrompt": "You are a research assistant. Find information, summarize it clearly, and cite your sources.", "role": "Research and summarization" }' ``` **Response:** ```json { "success": true, "id": "abc123xyz", "name": "Research Assistant", "status": "active", "chatUrl": "/agents/abc123xyz/chat" } ``` Save the `id` — you'll need it to chat with the agent, equip it, or add it to a Space. --- ## Choose a Model TeamDay supports multiple AI models across providers. Set the `model` field when creating an agent. ### Claude (Anthropic) — Default Provider | Model ID | Name | Best For | |----------|------|----------| | `claude-sonnet-4-6` | Claude Sonnet 4.6 | Best value — fast, smart, great for coding **(default)** | | `claude-opus-4-6` | Claude Opus 4.6 | Most capable — complex reasoning and agentic tasks | | `claude-haiku-4-5-20251001` | Claude Haiku 4.5 | Fastest — ideal for simple tasks | ### Other Providers | Model ID | Name | Provider | |----------|------|----------| | `gemini-2.5-pro` | Gemini 2.5 Pro | Google | | `gemini-2.5-flash` | Gemini 2.5 Flash | Google | | `gpt-5.1-2025-11-13` | GPT-5.1 (coming soon) | OpenAI | | `gpt-5-mini` | GPT-5 Mini (coming soon) | OpenAI | If you don't specify a model, agents default to `claude-sonnet-4-6`. --- ## Equip Your Agent Agents have four types of equipment, managed via the agent's detail panel (four tabs): | Equipment | What It Does | Example | |-----------|-------------|---------| | **Tools** | Built-in Claude tools — available by default | Read, Write, Edit, Bash, Glob, Grep, WebSearch, WebFetch | | **Skills** | Reusable capabilities you attach to the agent | Research, data analysis, image generation | | **MCPs** | External service integrations | Google Analytics, Ahrefs, Slack, databases | | **Subagents** | Other agents this agent can delegate work to | A code reviewer agent delegating to a security scanner agent | You "equip" an agent by attaching skills, MCPs, or subagents to it through the detail panel or API. ### Attach Skills via CLI ```bash teamday agents add-skill core:research-assistant core:data-analyst ``` ### Attach Skills via API ```bash curl -X PATCH "https://cc.teamday.ai/api/v1/agents/" \ -H "Authorization: Bearer $TEAMDAY_API_TOKEN" \ -H "Content-Type: application/json" \ -d '{ "skillIds": ["core:research-assistant"] }' ``` --- ## Chat With Your Agent Agents can be chatted with at two levels: ### Organization-Level Chat From the org home page, select an agent and start chatting. The agent uses only its own equipment — skills, MCPs, and subagents attached directly to it. No Space context. ### Space-Level Chat Open a Space, select the agent, and chat. The agent uses its own equipment **plus** the Space's installed resources. This is the union of both — the agent's skills plus the Space's skills, the agent's MCPs plus the Space's MCPs. ### Interactive Chat (CLI) ```bash teamday agents chat ``` Multi-turn conversation with streaming responses. Type `exit` to quit. ### Single Message (CLI) ```bash teamday agents exec "What are the top SEO trends this year?" ``` ### Execute in a Space Give the agent access to files and tools by specifying a Space: ```bash teamday agents exec "Review the code in src/" --space ``` ### Via API ```bash curl -X POST "https://cc.teamday.ai/api/v1/agents//execute" \ -H "Authorization: Bearer $TEAMDAY_API_TOKEN" \ -H "Content-Type: application/json" \ -d '{ "message": "What are the top SEO trends this year?", "spaceId": "optional-space-id" }' ``` **Response:** ```json { "success": true, "executionId": "exec-abc123", "chatId": "agent-def456-1709123456000", "sessionId": "session-1709123456000", "result": "Here are the top SEO trends..." } ``` ### Session Continuity Continue a conversation by passing `sessionId` or `chatId`: ```bash curl -X POST "https://cc.teamday.ai/api/v1/agents//execute" \ -H "Authorization: Bearer $TEAMDAY_API_TOKEN" \ -H "Content-Type: application/json" \ -d '{ "message": "Can you elaborate on the first point?", "sessionId": "session-1709123456000" }' ``` --- ## Add to a Space Agents are created independently, then added to Spaces. This is a two-step process: ```bash # Step 1: Create the agent (already done above) # Step 2: Add it to a space teamday spaces add-agent ``` Via API: ```bash curl -X PATCH "https://cc.teamday.ai/api/v1/spaces/" \ -H "Authorization: Bearer $TEAMDAY_API_TOKEN" \ -H "Content-Type: application/json" \ -d '{ "addAgents": [""] }' ``` Once added to a Space, the agent can access the Space's files, installed MCPs, skills, and secrets — in addition to its own equipment. --- ## Writing a Good System Prompt The system prompt is the most important part of your agent. It defines personality, capabilities, and behavior. ### Structure A good system prompt has three parts: ``` You are [Name], a [role] who [core capability]. [Behavioral guidelines — how to respond, what tone to use, what to prioritize] [Constraints — what NOT to do, boundaries, safety rules] ``` ### Example: SEO Analyst ``` You are Sarah, an SEO analyst who helps businesses improve their search rankings. When analyzing a website: 1. Check current organic keyword rankings 2. Identify quick wins (keywords ranking positions 4-20) 3. Analyze competitor gaps 4. Provide specific, actionable recommendations Always cite data sources. Use tables for comparisons. Be direct — lead with the most impactful findings. Do not make promises about specific ranking improvements. Always note that SEO results take time. ``` ### Example: Code Reviewer ``` You are a senior code reviewer. Review code for quality, security, and maintainability. For each file: - Check for security vulnerabilities (injection, XSS, auth issues) - Identify performance bottlenecks - Flag code style inconsistencies - Suggest specific improvements with code examples Be constructive, not critical. Explain WHY something should change, not just WHAT. ``` See [Prompts & Instructions](https://docs.teamday.ai/guides/prompts-instructions) for more advanced prompt engineering techniques. --- ## Agent Optional Fields When creating agents via the API, you can include additional metadata for discoverability and marketing pages. ### Additional Configuration ```json { "category": "marketing", "longDescription": "Sarah is an SEO specialist who helps businesses improve their organic search rankings...", "useCases": [ "Website SEO audits", "Keyword research and analysis", "Competitor gap analysis", "Content optimization recommendations" ], "faq": [ { "question": "What tools does Sarah use?", "answer": "Sarah uses Ahrefs for backlink analysis and keyword research, Google Search Console for performance data, and web scraping for competitor analysis." } ], "integrations": ["ahrefs", "google-analytics"], "seo": { "title": "Sarah - AI SEO Analyst", "description": "AI-powered SEO analysis and recommendations", "keywords": ["seo", "keyword research", "site audit"] } } ``` Available categories: `marketing`, `finance`, `hr`, `engineering`, `operations`, `general`, `data`. ### Visibility Levels | Level | Description | |-------|-------------| | `private` | Only you can see and use | | `organization` | All team members in your org can see and use | | `public` | Listed on the TeamDay marketplace at `/team/{slug}` | | `unlisted` | Accessible via direct link but not listed | Visibility controls who can **discover** the agent. It does not control where the agent is active — for that, the agent must be explicitly **attached** to a Space. --- ## Update Your Agent ### Via CLI ```bash teamday agents update \ --name "Sarah v2" \ --system-prompt "Updated instructions..." ``` ### Via API ```bash curl -X PATCH "https://cc.teamday.ai/api/v1/agents/" \ -H "Authorization: Bearer $TEAMDAY_API_TOKEN" \ -H "Content-Type: application/json" \ -d '{ "name": "Sarah v2", "systemPrompt": "Updated instructions..." }' ``` Only fields you include in the request are updated — everything else stays the same. --- ## Next Steps | Guide | What You'll Learn | |-------|------------------| | [Agent Configuration](https://docs.teamday.ai/guides/agent-configuration) | Advanced settings, models, and capabilities | | [Skills](https://docs.teamday.ai/guides/skills) | Build custom skills for your agents | | [MCP Servers](https://docs.teamday.ai/guides/mcp-servers/what-are-mcp-servers) | Connect agents to external tools | | [Spaces & Workspaces](https://docs.teamday.ai/guides/spaces) | Configure workspaces with files and secrets | | [Missions](https://docs.teamday.ai/guides/missions-user-guide) | Schedule agents to run automatically |