// the find
Fosowl/agenticSeek
Fully Local Manus AI. No APIs, No $200 monthly bills. Enjoy an autonomous agent that thinks, browses the web, and code for the sole cost of electricity. 🔔 Official updates only via twitter @Martin993886460 (Beware of fake account)
AgenticSeek is a locally-running autonomous agent that can browse the web, execute code, and manage files — positioned as a self-hosted alternative to Manus AI. It chains multiple specialized sub-agents (browser, coder, file, planner) with a routing layer that picks the right one for each task. Aimed at developers who want agent capabilities without cloud API costs or data leaving their machine.
The multi-agent routing architecture is well-thought-out — separate agents for browsing, coding, file ops, and planning keeps each agent's context focused rather than one bloated generalist. SearxNG as the search backend is a good call: no Google API key, no rate limits, and it stays local. The remote LLM server option (run the model on a beefy machine, use it from a laptop) shows real-world thinking about hardware constraints. Docker Compose setup makes the dependency stack (Redis, SearxNG, frontend, backend) reproducible without manual service wiring.
The hard requirement for Python 3.10.x specifically is a red flag — that version is approaching end-of-life and pinning to a minor version usually means dependency hell is hiding underneath. The `config.ini` approach for configuration is weak: no validation, comments break it, and mixing booleans/strings/URLs in a flat INI file without schema enforcement will cause subtle bugs (the README even warns about trailing spaces). The ChromeDriver version management is entirely manual and brittle — if Chrome auto-updates, things break silently, and the README's lengthy troubleshooting section for this is a sign the problem isn't solved, just documented. No test coverage for the core agent logic; the test suite covers parsing and providers but nothing about whether the agents actually complete tasks correctly.