Guide¶
A deep dive into what TBD Agents offers and how each capability works.
Prerequisites
Make sure you have a running instance first — see Quick Start.
-
Agents
Build custom AI agents with system prompts, models, and tool access.
-
Dashboard
Operate agents, workflows, credentials, guardrails, tasks, schedules, and chat from the Flutter UI.
-
MCP Tools
Connect external tools via the Model Context Protocol.
-
Skills
Reusable instruction modules that shape agent behaviour.
-
Workflows
Execution contexts that tie agents to prompts and sessions.
-
Streaming
Real-time SSE streaming of logs, messages, and usage.
-
Knowledge Bases
Connect vector DBs and MongoDB for retrieval-augmented generation.
-
Agent Memory
Use STM, LTM, and semantic memory for persistent agent context.
-
Guardrails
Enforce safety policies before agent execution begins.
-
BYOK Providers
Bring your own API keys for OpenAI, Azure, Anthropic, or custom endpoints.
-
Custom Tools
Write Python tool functions and lock them to specific agents.
-
Agent Chat
Chat with an agent over SSE without starting a workflow task.
-
Tasks
Run workflows, inspect task executions, and stop active runs.
-
Observability
Trace agent runs and monitor the application with OpenTelemetry.
-
Import & Export
Back up, migrate, and share agent configurations as JSON bundles.
-
PostgreSQL Backend
Run a fully-PostgreSQL stack (JSONB + pgvector) with no MongoDB dependency.