Not Human Search
** — Agent-first search engine indexing 8,000+ MCP servers and other agentically-readable sites, ranked across 7 signals (llms.txt, OpenAPI, ai-plugin, MCP, structured API, robots.txt, schema.org). Queryable via MCP (`nothumansearch.ai/mcp`), REST API, or browser. Includes `verify_mcp` live JSON-RPC probe tool. MIT.
Last verified:
What is Not Human Search?
Not Human Search is an agentic-first search engine built for AI agents rather than human web browsing. It indexes websites, APIs, tools, and services, then ranks them by how well they support non-human users and programmatic access.
The core idea is agentic readiness scoring: each indexed site gets a score from 0 to 100 based on seven signals such as llms.txt, ai-plugin.json, OpenAPI specs, structured APIs, MCP support, robots.txt AI rules, and Schema.org markup. The site describes score bands for Agent Ready, Partial Support, and Basic/None, so agents can quickly judge whether a target is usable.
It also exposes a full REST API and an MCP interface, so AI systems can search, inspect site details, check stats, and submit new sites at runtime. The homepage shows indexed examples, featured sites, and search prompts, which makes it useful both as a directory and as a discovery layer for agent-friendly services.
The product is aimed at developers, AI agent builders, automation teams, and businesses that want their services found and used by agents. It is especially relevant for anyone building tools that need to discover APIs, documentation, job boards, ecommerce services, or other machine-callable web services.
Not Human Search pricing
Pricing model: Freemium
The website presents the core search experience, REST API access, and MCP access as free to use. It explicitly says the remote MCP server is free, requires no account or authentication, and has no upsells or hidden costs. The site also offers free monitoring that emails you when your site’s agentic signals regress, and it lets you submit a site for indexing and scoring from the homepage or API. No paid plans or premium tiers are shown on the website.
Not Human Search pros
- Ranks sites by agentic readiness
- Built specifically for AI agents
- Indexes 1,900+ sites
- Single API call discovery
- Full REST API access
- MCP server support
- Searches APIs and services
- Scores sites from 0 to 100
- Uses seven concrete signals
- Checks llms.txt presence
- Checks ai-plugin.json support
- Checks OpenAPI specifications
- Detects structured API endpoints
- Recognizes robots.txt AI rules
- Supports Schema.org signals
- Shows featured agent-ready sites
- Offers site submission flow
- Provides site detail lookup
- Includes stats endpoint
- Useful for AI tool discovery
Not Human Search cons
- Not a general human search engine
- Coverage limited to indexed sites
- Quality depends on site metadata
- Low-score sites may be hard to compare
- Best for API/tool discovery, not broad web search
- Scoring can miss real-world usability
- Some signals are optional, not mandatory
- No evidence of advanced pricing tiers on site
- Primarily useful for technical users
Frequently asked questions about Not Human Search
What does Not Human Search do?
It is a search engine for AI agents that finds services, APIs, tools, and other web resources that are usable by non-human clients. Results are ranked by agentic readiness, so agents can prioritize sites that are easier to read, call, and integrate programmatically.
How does the scoring work?
Each indexed site gets a score from 0 to 100 based on seven signals. These include llms.txt, ai-plugin.json, OpenAPI specs, structured API endpoints, MCP server support, robots.txt AI rules, and Schema.org markup.
What does agentic readiness mean?
Agentic readiness is the measure of how well a website serves AI agents instead of human visitors. The score reflects whether the site exposes machine-readable docs, APIs, manifests, or other signals that make automated use easier.
How many sites are indexed?
The site states that it indexes 1,900+ sites, and the homepage shows a larger live count of indexed entries. It presents the index as a searchable directory of agent-ready websites and services.
Does Not Human Search have an API?
Yes. The site says it provides a full REST API with endpoints for search, site details, submission, and stats. It also points to /openapi.yaml and /llms.txt for documentation.
Does it support MCP?
Yes. The site offers an MCP server so agents can call functions such as search_agents, get_site_details, get_stats, and submit_site. The homepage also shows connection instructions for MCP-compatible clients.
Can I submit my own site?
Yes. The website lets you submit a site from the homepage or by POSTing to the submission API. After submission, the site can be crawled, scored, and included in agent search results.
What are the score ranges?
The site groups scores into three bands: 70 to 100 is Agent Ready, 40 to 69 is Partial Support, and 0 to 39 is Basic or None. These ranges help agents quickly decide whether a site is likely to work well.
Is there a free monitoring feature?
Yes. The homepage says you can set up free monitoring and receive an email when your site’s agent signals regress. It is limited to one email per week and only sends alerts for real regressions.
Who is this for?
It is for AI agent builders, automation developers, and businesses that want their services discoverable by machine users. It is also useful for teams that want to improve documentation, APIs, and other signals that make sites easier for agents to use.