Manifest
Manifest turns any URL into a structured JSON map of actionable UI elements—buttons, forms, and inputs—so AI agents can reliably click, fill, and submit without fragile selectors or screenshots.
https://omfang.io/?ref=producthunt
Product Information
Updated:Aug 8, 2026
What is Manifest
Manifest (by omfang) is an “action layer” API for AI agents that converts a webpage into a clean, agent-readable representation of what can be done on that page. Instead of only showing an agent what a page looks like (browser tools) or what it says (scrapers/content extractors), Manifest identifies the interactive controls—such as inputs, buttons, and navigational links—and describes them with the context an agent needs to act correctly, including what’s required before it can proceed.
Key Features of Manifest
Manifest is an API that turns any webpage URL into a structured JSON “action map” an AI agent can reliably reason about and execute. It analyzes the page’s accessibility tree (like a screen reader) and cross-checks it with the underlying DOM to capture actionable UI elements—buttons, forms, inputs—and the details agents need to succeed (required fields, input types, placeholders, disabled states), avoiding fragile CSS selectors and screenshot-guessing while remaining resilient to UI redesigns.
URL → JSON action manifest: One API call converts a webpage into a clean, structured list of actions (click/fill/submit) plus navigation links, making pages programmatically operable for agents.
Accessibility-tree driven understanding: Reads the accessibility tree the way assistive tech does, producing stable, semantic descriptions of interactive elements instead of brittle visual or selector-based heuristics.
DOM cross-referencing for missing details: Augments accessibility data with DOM-derived specifics such as required fields, input types, placeholders, and disabled states—critical for completing forms correctly.
Agent-reasonable output (not raw HTML): Returns actions with labels, types, descriptions, and requirement flags so agents can plan steps (e.g., what must be filled before submitting) without custom parsing.
Resilient to UI redesigns: Avoids hand-maintained selectors and fragile CSS paths that break when pages change, improving long-term reliability for automation.
Caching for speed and predictable cost: Cached responses are fast and intended to reduce repeated costs for agents that revisit the same URLs frequently.
Use Cases of Manifest
AI agent web task automation: Enable agents to reliably complete multi-step website tasks (fill forms, click buttons, submit flows) using a structured action layer rather than screenshots or brittle selectors.
Customer support and self-serve workflows: Power support agents/bots that can navigate account portals, submit requests, and guide users through web forms by understanding what actions are available and required.
Sales and lead capture automation: Automate outreach operations like submitting contact forms, requesting demos, or completing partner onboarding forms while correctly handling required fields and input types.
QA and regression testing for web apps: Generate a stable inventory of interactive elements for test automation that is less sensitive to UI refactors, improving maintainability of end-to-end tests.
RPA modernization for enterprises: Replace selector-heavy RPA scripts with a semantic action manifest layer to reduce breakage across internal tools and third-party web systems.
Accessibility-aware automation: Leverage the accessibility tree to drive automations that align with accessible semantics, helping ensure agents interact with pages in a robust, screen-reader-like way.
Pros
More robust than hand-written selectors by relying on accessibility semantics and DOM validation.
Agent-friendly structured JSON output that explicitly lists available actions and requirements.
Avoids screenshot-based guessing, enabling more deterministic automation flows.
Caching can improve performance and make repeated interactions more cost-predictable.
Cons
Effectiveness may depend on the quality of a site’s accessibility tree and DOM structure (poorly labeled pages may yield less useful manifests).
Primarily describes what can be done on a page; executing actions still requires an agent runtime/browser tool.
Dynamic or highly scripted UIs may require additional handling beyond static element discovery.
How to Use Manifest
1) Get access to Manifest: Go to https://omfang.io/ and click “Get your API key” (or visit https://app.manifest.omfang.io/login) to sign in and obtain an API key for the Manifest API.
2) Decide which page (URL) you want to map into actions: Pick the exact URL your agent needs to operate on (e.g., a login page, checkout page, contact form). Manifest works by visiting a URL and producing a structured action map for that page.
3) Call the Manifest endpoint to generate an action manifest: Send a request to the Manifest API endpoint shown in the docs: POST /manifest with a JSON body like { "url": "https://example.com" }. This triggers Manifest to visit the page, read the accessibility tree, cross-reference the DOM, and return a structured JSON manifest of available actions.
4) Read the returned page summary (current_page_state): In the response JSON, use the "current_page_state" field to understand what page you are on (e.g., “Contact page with a message form”). This helps an agent confirm context before taking actions.
5) Enumerate available actions (buttons, forms, inputs): Inspect the "actions" array in the response. Each action includes fields like id, label, type, description, and required. This is the core output: a structured list of what an agent can click, fill, or submit.
6) Identify required fields and constraints before acting: Use each action’s metadata (e.g., "required": true/false) to determine what must be filled in before submission. Manifest is designed to include details accessibility trees often miss, such as required fields, input types, placeholders, and disabled states.
7) Use navigation links to move between pages when needed: Inspect the "navigation" array (e.g., { "label": "Home", "url": "/" }) to find stable navigation targets. Use these URLs to guide the agent to the next page, then re-run POST /manifest on the new URL to get the next page’s action map.
8) Iterate: re-run Manifest after each page change: Whenever the agent navigates to a different page or the UI state changes, call POST /manifest again for the current URL. This keeps the agent’s action map aligned with the live page state without relying on fragile selectors or screenshots.
9) Try it live before integrating: Use the live demo at https://demo.manifest.omfang.io/demo to see the concept end-to-end: URL in → structured JSON action manifest out.
10) Manage usage and cost with caching and a plan: Choose a plan from https://omfang.io/#pricing based on expected call volume (Free/Starter/Pro/Enterprise). Note that cached responses are described as fast and typically don’t count against rate limits the same way cold requests do, which helps keep repeated agent workflows predictable.
Manifest FAQs
Manifest is an “action layer” API that turns any URL into a structured JSON map of what an AI agent can do on a webpage—buttons, forms, and inputs—plus what’s required before an agent can proceed.
Manifest Video
Popular Articles
Atoms: A Multi-Agent AI Platform That Transforms Ideas into Launch-Ready Products
May 22, 2026
Nano Banana SBTI: What It Is, How It Works, and How to Use It in 2026
Apr 15, 2026
Atoms Review — The AI Product Builder Redefining Digital Creation in 2026
Apr 10, 2026
Kilo Claw: How to Deploy and Use a True "Do‑It‑For‑You" AI Agent(2026 Update)
Apr 3, 2026
Latest AI Tools Similar to Manifest
Jorpex
Jorpex is a comprehensive tender notification platform that aggregates and delivers instant tender alerts from across European countries directly to Slack, helping businesses never miss opportunities.
Leadsmrt
Leadsmrt is a lead generation tool that helps businesses scrape, verify and personalize targeted business leads from Google Maps with AI-powered personalization capabilities.
Omnial AI
Omnial AI is a data intelligence platform that leverages AI agents to transform web prompts into structured, actionable data insights backed by Afore Capital.
SERPrecon
SERPrecon is an advanced SEO tool that leverages vectors, machine learning, and natural language processing to help users analyze and outrank competitors by using the same methods as modern search engines.
Popular AI Tools Like Manifest
Magentic-UI
Magentic-UI is an open-source, human-centered web agent prototype by Microsoft Research that enables collaborative planning and execution of complex web-based tasks through transparent interaction between humans and AI agents.
Web Accessibility Testing MCP
A11y MCP is a Model Context Protocol server that enables AI language models to perform automated web accessibility testing by integrating with Deque's Axe-core API and Puppeteer to analyze web content for WCAG compliance.
Open Browser Use
Open Browser Use is an open-source, agent-runtime-neutral browser automation layer that pairs a Chrome extension with a CLI/SDK/MCP to enable DOM-aware, CDP-powered tab control, navigation, and actions across different AI agent tools.
Web Researcher MCP
Web Researcher MCP is a local, production-grade MCP server that lets AI assistants search the web (optionally restricted to trusted “search lenses”), read full sources (including JS-rendered pages, PDFs, and YouTube transcripts), and return verifiable citations with multi-provider failover. [Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt]