Manifest
Overview
What is Manifest? Manifest is a service that turns any webpage into a structured action manifest for AI agents. Instead of returning raw HTML or a screenshot, it returns JSON describing what's actually clickable, fillable, and submittable on the page — buttons, inputs, forms, and navigation — so an agent can act on a page instead of just reading it.
How to use Manifest? Get an API key from app.manifest.omfang.io, then call the get_manifest tool with a URL. The server returns a structured manifest describing the page's actions. It also works as a Python SDK (pip install manifest-api) and has a LangChain tool integration for building agents directly.
Key features of Manifest?
- Returns clickable, fillable, and submittable actions — not just page content
- Detects required fields, input types, and disabled elements that standard accessibility trees miss
- Caches manifests, so repeat calls to the same URL are near-instant
- Works as an MCP server, a Python SDK, or a LangChain tool
Use cases of Manifest?
- Letting an AI agent fill out and submit a web form without hardcoded selectors
- Giving an agent a reliable way to navigate a site whose UI changes frequently
- Building browser-automation agents that need to reason about what actions are available, not just what text is on the page
FAQ from Manifest?
-
What's the difference between Manifest and a browser automation tool like Browserbase?
-
Browserbase gives an agent access to a browser. Manifest gives an agent a structured understanding of what it can do on a page it's already looking at — the two are complementary, not competing.
-
Does Manifest require an API key?
-
Yes, get one at app.manifest.omfang.io. A free tier is available.
-
What does the response look like?
-
A JSON object with the page's current state, a list of actions (with type, label, description, and whether each is required), and navigation links.
Server Config
{
"mcpServers": {
"manifest": {
"url": "https://manifest.omfang.io/mcp"
}
}
}