Build & orchestrate
Agentkit
Model-agnostic agents, tools, multi-agent graphs and swarms, MCP, streaming, and sessions.
Build agents, compose Core building blocks, manage prompts, enforce guardrails, and observe LLM systems — one documentation home for the full stack.
Products
Pick the product that matches your job — each has its own guides, APIs, and Get Started path.
Build & orchestrate
Model-agnostic agents, tools, multi-agent graphs and swarms, MCP, streaming, and sessions.
Compose building blocks
LLMs, embeddings, vector stores, retrieval, memory, Graph RAG, document processing, and integrations.
Version & release prompts
Author, review, version, and promote prompts across environments. Fetch the active version at runtime via SDK.
Protect & enforce policy
Toxicity, PII/PHI, token budgets, tool authorization, rate limiting, and programmable input/output rails.
Observe & measure
Track usage, cost, latency, RAG, and agent traces. Project-based observability for SaaS and on-prem.
Platform
Use them independently, or as one path from build to production observability.
Orchestrate agents and Core components.
Ship reviewed prompt versions without redeploys.
Run safety and policy checks before and after the LLM.
Measure cost, latency, and traces in production.
Ecosystem
Model hosts, vector stores, cloud storage, databases, OCR, and protocol integrations used across Core and Agents.