TeamWeb AI is a self-hosted platform for building intelligent AI assistants. Create assistants with distinct personalities and capabilities, give them access to your knowledge, and deploy them across the channels your team already uses.
Intelligent Assistants
Each assistant has its own personality, instructions, and capabilities. Choose from four operating modes depending on what you need:
Open mode for general-purpose assistants that can discuss any topic and use any tool. Knowledge-only mode for support agents that must answer strictly from your documentation — no guessing, no hallucination. Guided modes for structured interviews and discovery flows that collect information naturally before triggering actions.
Assistants can chat, search knowledge bases, generate content, execute code in sandboxed environments, run automated tasks, and communicate via email — all configured through a straightforward web interface.
Knowledge That Stays Current
Build knowledge bases from multiple sources — crawl websites, upload PDFs and Office documents, index GitHub repositories, or add structured facts directly. TeamWeb AI uses vector embeddings for semantic search, so assistants find relevant information even when the wording doesn’t match exactly.
Knowledge sources auto-sync on daily or weekly schedules, detecting content changes automatically so your assistants always have up-to-date information.
Connect Everywhere
Deploy assistants across seven channels: browser chat for your team, Slack and Telegram for messaging, email via Postmark, voice conversations via WebRTC, API triggers for connecting to external automation tools, and an embeddable widget you can drop onto any website with a single script tag.
The widget uses Shadow DOM for complete style isolation, supports lead capture, persists conversations across page navigation, and includes rate limiting out of the box.
Tasks, Schedules, and Automation
Define reusable tasks with specific objectives and output formats — blog posts, reports, summaries, spreadsheets, or images. Run them manually, on a cron schedule, or trigger them from external services via webhooks.
Generated content goes through a review workflow: assistants produce drafts, submit them for human review, and respond to approval or rejection feedback automatically. Blog posts can publish directly to WordPress.
Tools and Extensibility
Assistants come with built-in tools for web search, code execution, note-taking, and content creation. Extend them with custom HTTP tools from any API or OpenAPI spec, Python tools written in a built-in editor, MCP servers (Docker-based or external), or plugins.
The plugin system supports four types: tool plugins that add new capabilities, channel plugins for new communication platforms, provider plugins for additional LLM backends, and output plugins for custom integrations. Code execution runs in sandboxed Docker containers for security.
Built for Teams
Organise work into projects with shared knowledge bases, content guidelines, and plugins. Multiple assistants can collaborate within a project, sharing state and coordinating through a key-value store. Manage users, track conversations, monitor API usage, collect feedback, and configure notifications and email digests — all from a single admin interface.