Manifest
Open-source LLM gateway that keeps your AI agents running
8 followers
Open-source LLM gateway that keeps your AI agents running
8 followers
Manifest is a simple, open-source LLM gateway for your AI agents and apps. Your agents send every model call to one endpoint, and you set the routing without touching your code. Manifest tracks every request, token and dollar, and keeps your agents running with fallbacks and self-healing. Manifest works with every major provider, your own keys and subscriptions, local models such as Ollama or LM Studio, and custom providers. Start free on Manifest's cloud or self-host Manifest with Docker.
Free
Launch Team / Built With
AWS Startup AdvisorPersonalized AWS guidance that meets you where you’re already building
Promoted
👋 Hey PH!
Today we’re launching Manifest, a simple observability platform for OpenClaw agents. It gives you visibility into what each agent is costing you so you can track token usage, monitor spending across your OpenClaw agents, and get notified when a budget limit is reached, helping you stay in control before costs spiral out of hand.
The problem
OpenClaw doesn’t clearly show what your agents are actually costing you in real time. It is hard to understand which agent is consuming the budget, which message caused a spike, or which model is driving usage. This visibility gap is where costs spiral.
What we built
Manifest gives you live cost and token observability across all your OpenClaw agents. You can see cost globally or per agent, message or model. Instead of guessing, you can identify exactly what is burning your budget and fix it.
Simplicity
Manifest is designed to be simple and fast to set up. No complex configuration. No heavy setup. Just clear visibility in seconds.
Privacy
Manifest is open source and fully self-hostable. You can deploy it on your own infrastructure and keep everything under your control. We do not collect prompts or message content. Only telemetry metadata such as tokens, model name, action type, and timestamps.
How it works
Manifest connects via OpenTelemetry with one ingest key per agent. No prompt scraping. No middleware intercepting your conversations. Just structured telemetry you choose to send.
Who it’s for
Anyone running an OpenClaw agent who wants to control costs easily instead of discovering them at the end of the month.
It’s still early and we’re building in public. I’d genuinely love feedback from the OpenClaw community.