How does Manifest handle provider outages or rate limits?

It supports up to five fallback models per tier to keep requests alive when a provider rate-limits or returns a 5xx error. An auto-fix layer also rewrites deprecated model IDs and malformed parameters before they hit the upstream API.

Can I use my existing AI subscriptions with Manifest?

Yes, Manifest allows you to connect paid subscriptions from providers like OpenAI, Anthropic, and Gemini, as well as local runtimes like Ollama and LM Studio. This enables routing calls to models you are already paying for.

Is Manifest compatible with existing agent SDKs?

Manifest provides an OpenAI-compatible endpoint, meaning most SDKs and agent harnesses can drop in unchanged. You simply point your agent or SDK at Manifest's endpoint to start routing traffic.

What are the deployment options for Manifest?

The project is fully open source with a self-hosted Docker path available. There is also a managed cloud offering currently in early access, though production users are mostly on self-hosted instances.

Who is Manifest designed for?

It is aimed at teams running real agent workloads who want budget caps, spend visibility, and the ability to route cheap calls to cheaper models. It is not recommended for single-provider applications that do not need routing or fallback chains.

Manifest

Open-source LLM router that fans your agent traffic across providers and your existing AI subscriptions.

Freemium· Free: $0 /month · Pro: $19 /month · Enterprise: Let's TalkAgentsMulti-model7.1 / 10

Visit website →

In short

Manifest is an open-source routing layer that sits between agents and LLM providers to manage traffic, costs, and reliability. It is best for teams running agent workloads who need a single endpoint with fallbacks and budget controls without rewriting their application code.

Best for

Pick Manifest if you run agent workloads across multiple LLM providers and want a single OpenAI-shaped endpoint with budgets, fallbacks and auto-fix in front of them.

Skip if

Skip it if you're calling one provider from one app and don't need routing, budget caps or fallback chains.

Manifest is an OpenAI-compatible routing layer that sits between your agents and the LLM providers behind them. You point your agent or SDK at Manifest's endpoint, connect provider keys or paid subscriptions (OpenAI, Anthropic, Gemini, plus local runtimes like Ollama, Llama.cpp and LM Studio), and define rules that decide which model handles which call. Up to five fallback models per tier keep requests alive when a provider rate-limits or 5xxs, and an auto-fix layer rewrites deprecated model IDs and malformed parameters before they hit the upstream API.

It's aimed at teams running real agent workloads (Claude Code, Hermes, OpenClaw and similar harnesses are called out by name) who want budget caps, spend visibility, and the ability to route cheap calls to a cheap model without rewriting their app. The project is fully open source with a self-hosted Docker path (48k+ pulls, 7k+ GitHub stars at time of writing) plus a managed cloud offering in early access. BYOK keeps you in control of vendor billing while Manifest handles the orchestration.

Think of it as a more agent-focused alternative to LiteLLM or OpenRouter: less of a marketplace, more of a policy engine for inference you're already paying for.

Editor's take

Manifest is the kind of glue layer agent builders eventually write themselves, except it's already open source and battle-tested by thousands of self-hosters. The auto-fix and subscription-routing angles are genuinely novel; most competitors stop at simple proxying. We'd self-host it before trusting the early-access cloud.

— The AI Tool Bible editorial team

Pros

✅ OpenAI-compatible endpoint, so most SDKs and agent harnesses drop in unchanged
✅ Per-tier fallbacks keep agents running through provider outages and rate limits
✅ Auto-fix silently corrects deprecated model IDs and bad parameters
✅ Self-hostable via Docker with support for local runtimes (Ollama, LM Studio, Llama.cpp)
✅ BYOK plus subscription routing lets you actually use the ChatGPT/Claude plans you already pay for

Cons

⚠️ Adds a network hop and another moving piece to debug when calls fail
⚠️ Managed cloud is still early access; production users are mostly on self-host
⚠️ Routing rules and budgets only help if you actually configure them thoughtfully

Use cases

llm-routingcost-controlagent-infrastructurefallback-handlingmulti-provider

Frequently asked

How does Manifest handle provider outages or rate limits?: It supports up to five fallback models per tier to keep requests alive when a provider rate-limits or returns a 5xx error. An auto-fix layer also rewrites deprecated model IDs and malformed parameters before they hit the upstream API.
Can I use my existing AI subscriptions with Manifest?: Yes, Manifest allows you to connect paid subscriptions from providers like OpenAI, Anthropic, and Gemini, as well as local runtimes like Ollama and LM Studio. This enables routing calls to models you are already paying for.
Is Manifest compatible with existing agent SDKs?: Manifest provides an OpenAI-compatible endpoint, meaning most SDKs and agent harnesses can drop in unchanged. You simply point your agent or SDK at Manifest's endpoint to start routing traffic.
What are the deployment options for Manifest?: The project is fully open source with a self-hosted Docker path available. There is also a managed cloud offering currently in early access, though production users are mostly on self-hosted instances.
Who is Manifest designed for?: It is aimed at teams running real agent workloads who want budget caps, spend visibility, and the ability to route cheap calls to cheaper models. It is not recommended for single-provider applications that do not need routing or fallback chains.