AutoGen
Microsoft's multi-agent conversation framework (now in maintenance mode).
TypeScript multi-agent networks with deterministic routing and MCP
AgentKit's edge is its pairing with Inngest's durable execution engine, which addresses a real pain point: making long-running, multi-step agent workflows fault-tolerant and retryable in production. Deterministic routing appeals to teams burned by unpredictable, model-driven control flow. The trade-off is coupling to the Inngest ecosystem, which is great if you already use it and a bigger commitment if you don't. A strong fit for TypeScript teams wanting reliable, event-driven agent orchestration.
Inngest AgentKit is an open-source TypeScript library for building multi-agent networks with deterministic routing, MCP tooling, and Inngest-powered durable, fault-tolerant orchestration.
AgentKit is Inngest's TypeScript library for composing AI agents, from a simple single agent to entire networks in which multiple agents collaborate. Agents are stateless entities with a goal and optional tools; they are combined into Networks with a Router that decides which agent runs next, while shared Network State records typed conversation and workflow context. A core differentiator is deterministic and flexible routing: rather than leaving control flow entirely to the model, AgentKit lets developers define reliable routing logic, combining a fully typed state machine with conversation history. It works with multiple model providers and embraces MCP for rich tooling. Because it is built on Inngest, AgentKit benefits from the Inngest Dev Server for local development and Inngest's durable orchestration engine in production, making agent workflows fault-tolerant, retryable, and observable when deployed to the cloud.
Inngest AgentKit is an open-source TypeScript agent framework whose edge is deterministic routing plus Inngest-powered durable, fault-tolerant orchestration.
AgentKit is developed by Inngest, a company known for durable execution and event-driven workflow orchestration. AgentKit extends that engine to AI agents, targeting TypeScript developers.
By building on Inngest, AgentKit inherits durability, retries, and observability, differentiating it from frameworks that leave production reliability to the developer.
AgentKit models agents as stateless goal-driven entities with tools, composed into Networks with a Router and typed Network State. It supports multiple model providers and MCP tooling, and offers deterministic routing via a state machine.
The Inngest Dev Server enables local iteration, while deployment on Inngest provides durable, fault-tolerant execution in the cloud.
TypeScript engineering teams building multi-agent systems that need reliable, durable orchestration, especially those already using or open to Inngest.
TypeScript developers building agent workflows
Engineering leads adopting durable orchestration
Backend and platform engineers
TypeScript teams needing durable, event-driven multi-agent systems
Inngest is a venture-backed company; verify specific funding details with the vendor.
The AgentKit library is open-source and free. Running durable workflows uses Inngest, which has free and paid tiers.
AgentKit lets you define routing logic with a typed state machine rather than relying solely on the model to decide control flow.
Yes. AgentKit embraces MCP for rich tooling across agents.
You can use the library standalone, but durable, fault-tolerant execution comes from deploying on Inngest.
AgentKit is a TypeScript library.
Side-by-side pages for pricing, features, and best-fit use cases.
Microsoft's multi-agent conversation framework (now in maintenance mode).
High-performance Python framework for building multi-agent systems and AgentOS
Graph-based orchestration for stateful, controllable LLM agents.
The TypeScript framework for building AI agents and workflows.