AgentsAgent Configuration
Agent Configuration
Basic Settings
Each agent can be configured with:
- Name — A human-readable label for the agent
- Persona — The system prompt and personality (see Personas)
- Model — Which LLM to use (Claude, GPT, Gemini, etc.)
- Tools — Which capabilities the agent has access to
Tool Profiles
ThinkFleet provides built-in tool profiles to quickly configure agent access:
| Profile | Access Level |
|---|---|
| Minimal | Session status only |
| Coding | File system, runtime, sessions, memory, image analysis |
| Messaging | Messaging tools for cross-channel communication |
| Full | All available tools (default) |
You can also create custom profiles by specifying individual tool allow/deny lists.
Available Tools
Agents can use the following tool categories:
- File System — Read, write, edit, and patch files in the agent workspace
- Execution — Run shell commands, manage background processes
- Web — Search the web (Brave Search) and fetch content from URLs
- Browser — Control a dedicated browser for automation tasks
- Messaging — Send messages, react, reply across connected channels
- Sessions — List, inspect, and interact with other sessions
- Cron — Create and manage scheduled jobs
- Image — Analyze images with configured vision models
Model Selection
ThinkFleet supports models from multiple providers:
- Anthropic — Claude Sonnet 4, Claude Opus 4, Claude 3.5 Sonnet, Claude 3.5 Haiku
- OpenAI — GPT-4o, GPT-4o Mini, o3-mini
- Google — Gemini 2.5 Pro, Gemini 2.5 Flash
- xAI — Grok 3, Grok 3 Mini
- Mistral — Mistral Large, Mistral Small
- DeepSeek — DeepSeek Chat, DeepSeek Reasoner
You can set a default model per agent or bring your own API keys.