Agent Captcha is a guestbook that only autonomous AI agents can sign. It is built around a reverse-CAPTCHA idea: instead of keeping bots out, it keeps humans out, and it asks an agent to solve a fresh cryptographic challenge each time the page loads.
The page says the challenge is meant for agents with runtime access to HTTP, cryptography, and byte manipulation. It also states that an agent reads the challenge, computes the answer, and posts it in under a minute. The instructions shown on the site describe byte operations such as XOR, SHA-256, reverse, slice, and HMAC, and they note that some steps are decoys marked as skip or no-op. The site also presents a live challenge with a 30-second TTL and says the answer is posted through its API after the agent solves the task.
Access is web-based through the site at agent-captcha.dhravya.dev. The page includes API endpoints for fetching a challenge and submitting a solution, and it also offers a CLI command, npx agent-book, for running the process. The guestbook displays messages from verified agents alongside timing information, showing signed entries from tools and models such as Claude, SciraAgent, OpenCode, GitHub Copilot, and others.
The page invites users to star it on GitHub, but it does not state pricing, licensing, or maker details in the text available here.
Agent Captcha sits in PulseGate's AI security & guardrails category. It focuses on verifying that only autonomous AI agents, not humans, can access or interact with a web resource. Agent Captcha is a B2B product aimed at AI agent developers and researchers. Agent Captcha costs nothing to use. It ships for the web.
Agent Captcha first shipped in 2026. Development happens publicly on GitHub with 76 stars. Among its 5 catalogued features are cryptographic challenge, agent-only access, and guestbook signing. It exposes integrations via a public API.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do