Manus AI visits classified as unknown

Summary

Manus AI is an autonomous AI agent platform that browses the web on behalf of users. Its crawler identifies itself with a Manus-User/1.0 token in the user-agent string. Kaistone Radar’s bot detection patterns do not include a match for this token, so Manus AI visits are classified as “Unknown / Human” despite being automated AI agent traffic.

Background

The BOT_PATTERNS array in beacon.mjs contains regular expressions matched against the user-agent header. When a hit arrives, the patterns are tested in order and the first match determines the bot label. If no pattern matches and the IP is not in a known infrastructure range, the hit is labelled “Unknown / Human.”

At the time of this finding, the pattern list includes 30+ entries covering OpenAI, Anthropic, Google, Bing, Meta, and various HTTP client libraries — but Manus AI is absent.

Observation

On 25 March 2026, a hit was recorded with the following attributes:

The user-agent string appends Manus-User/1.0 after a standard Chrome UA, following the common convention of bot identification via UA suffix. The IP resolves to AWS infrastructure in the us-east-1 region, consistent with a cloud-hosted AI agent rather than a human visitor.

Notably, Manus AI visited the /tree/ root page, suggesting it followed the link from the home page. However, it did not proceed deeper into the tree structure — no hits were recorded for any child pages.

About Manus AI

Manus is an AI agent platform that performs multi-step web tasks autonomously. Unlike traditional search engine crawlers that index pages for later retrieval, Manus operates in real time — browsing, reading, and acting on web content as part of user-initiated tasks. This makes it a distinct category of AI web visitor: not a crawler in the traditional sense, but an autonomous agent that interacts with sites on behalf of its users.

Implications

The AI agent landscape is expanding rapidly. New platforms like Manus, Devin, and others are deploying web-browsing agents that visit sites with custom user-agent strings. Each new platform requires a corresponding pattern in the detection list. Without proactive pattern updates, an increasing proportion of AI traffic will be misclassified.

This highlights a structural challenge: the bot detection approach relies on a manually maintained list of known patterns. As the number of AI agent platforms grows, maintaining this list becomes a continuous effort. A community-maintained registry of AI bot user-agent patterns would benefit projects like Kaistone Radar.

Proposed Fix

Add a Manus pattern to the BOT_PATTERNS array:

{ pattern: /manus/i, name: 'Manus AI' }

Additionally, add a corresponding colour entry in the dashboard:

'Manus AI': '#e040fb'  // purple-pink

Longer term, consider implementing a more scalable approach to bot detection — such as loading patterns from a configurable JSON file or fetching from a shared registry — to reduce the friction of adding new AI platforms as they emerge.

Resolution

References

Beacon source: netlify/functions/beacon.mjs
Dashboard: /dashboard/
Findings index: /findings/