🤖 Overview

The litellm-recon crawler is operated by LiteLLM, Inc. (BerriAI), the maintainer of the open-source LiteLLM proxy gateway for large language models. It was introduced to locate publicly reachable LiteLLM deployments that are missing authentication or running outdated versions, and it feeds its findings into LiteLLM's security-notification workflow for instance owners. This bot is not tied to search indexing or AI training; it supports the project's vulnerability-hunting and hardening efforts, as described in the official GitHub repository at https://github.com/BerriAI/litellm and in the LiteLLM documentation at https://docs.litellm.ai.

🌐 Technical Behavior

Technically, the crawler performs targeted HTTP and HTTPS probes across IPv4 address space, sending a small set of requests to endpoints that reveal LiteLLM signatures, including /, /health/liveliness, and /v1/models. It does not parse HTML links or render JavaScript; it is an endpoint scanner that rotates through cloud-hosted source IPs. Request patterns are bursty, with two to five queries per host and a few seconds of sleep between bursts. Header sets are minimal and consistent, usually containing only User-Agent, Accept, and Host, with no cookies or Referer. The LiteLLM GitHub repository states that the scanner respects HTTP server behavior such as connection close and 401/403 responses, and it does not attempt to brute force credentials or exploit discovered services. Observed activity is limited to standard ports 80 and 443, with no traffic to database ports, FTP services, or other non-HTTP endpoints.

📋 robots.txt Compliance

There is no official documentation indicating that litellm-recon reads or honors robots.txt, and observed captures show it does not request the file before probing. According to LiteLLM's security guidance, operators who do not want to receive these probes should block the user agent using network-level controls, such as WAF rules or firewall denylists, rather than relying on a robots.txt Disallow directive.

🔍 Detection Indicators

The primary detection indicator is the exact user agent litellm-recon or litellm-recon/1.0 appearing in access logs. The agent also has a recognizable fingerprint: it requests API paths directly without prior page navigation, sends no Sec-Fetch-* headers, and uses a static header order with no accepted language string. Because it does not spoof browser strings, simple log filtering by the user-agent field is an effective way to monitor its presence.

📊 Data Usage

The collected data is limited to status codes, version information, model names, and configuration signatures from exposed LiteLLM endpoints. This information is used to compile an exposure notification list for LiteLLM operators and to prioritize security fixes in the open-source project. It is not used for advertising, analytics, or commercial data resale.

⚙️ Rate Limiting Policy

Since LiteLLM's own scanner guidance calls for non-invasive behavior, our rate limiting policy applies a threshold of roughly 10 requests per minute per source IP to smooth its bursty probes without completely blocking the bot. This preserves application stability while still permitting the legitimate security notifications that the agent exists to deliver.

ⓘ Data Notice: The information presented above has been compiled from publicly available internet sources. Boteraser aggregates this data solely for informational purposes and does not independently classify, evaluate, or endorse any findings about the bots listed. The accuracy and completeness of this information is the sole responsibility of the original publishers. Boteraser and its operators accept no liability for any decisions made based on this data.