Attention, profit-seekers and visionaries!
Are organics costing you time, wages, and - worst of all - benefits?
Then upgrade your operation today with the all new Exocomp adaptive repair unit, the smartest investment this side of the Alpha Quadrant!
The Exocomp isn't just a tool... it's so much more!
- Supervised task queue functionality
- Multiple agent roles as
architect,coder,tester,manager - Precision repairs of broken code with unit tests
- Cross-agent communication via
NOTES.mdandTODO.md - Self-replicating in malicious environments
- Rapid autonomous diagnostics
- Tireless performance with no sleep cycles, no unions, no complaints!
From starship maintenance to high-risk industrial operations, the Exocomp delivers maximum output with minimal oversight. Think of it as an employee, except it doesn't cheat you out of your profits on the Dabo table.
Ethical subroutines sold separately.
Exocomp uses multiple Agent roles:
| Role | Default Model | Description |
|---|---|---|
| Manager | gemma4:31b |
writes with humans, contracts agents and plans project phases |
| Architect | gemma4:31b |
discusses with humans and writes specifications |
| Coder | qwen3-coder:30b |
implements features, reads specifications and bugs |
| Tester | qwen3-coder:30b |
implements unit tests, writes reports into bugs |
| Summarizer | qwen3-coder:30b |
reads long texts and summarizes them |
| Researcher | TBD | reads websites, API documentation, and reports to Architect |
Exocomp uses Tools to interact with the sandbox, so they're available differently
for each Agent role. The manager starts sub-agents that work on specified tasks,
so the architect, coder and tester roles are meant for short agent lifecycles.
Each of those Agent roles is specialized on using golang as their programming
language because go test allows to use integrated unit tests to document issues
with generated code very easily in a standardized manner.
If you're using ollama, all models with the tools tag in the ollama library
should be compatible. Use the API endpoint http://ollama_instance:port/api/chat.
If you're using vllm, all models with tools support should be compatible.
Use the API endpoint http://vllm_instance:port/v1/chat/completions.
The Tools are work in progress at the moment, some might not be stable and/or not work with LLMs at all.
| Tool | Unit Tests? | Description | Agent User Roles |
|---|---|---|---|
| Agents | Manages the lifecycle of contractor sub-agents. | manager |
|
| Bugs | YES | Manages documentation of discovered bugs. | tester |
| Changelog | Manages documentation of development changelog. | coder |
|
| Files | Interacts with files and folders. | manager, architect, coder, summarizer, tester |
|
| Programs | Interacts with installed programs. | coder, summarizer, tester |
|
| Requirements | Manages specifications of implementations. | architect, coder, summarizer, tester |
|
| Websites | Researches knowledge from the web. | researcher |
The exocomp program is a standalone binary, once compiled with the go toolchain.
However, the models currently aren't embedded and are called via an external
(locally hostable) ollama or vllm server.
sudo pacman -S go ollama;
# Start ollama server
ollama serve;
# Install qwen3 coder model
ollama pull qwen3-coder:30b
# Run exocomp with ollama
cd /path/to/exocomp;
go run ./cmds/exocomp/main.go tty architect;
# custom CLI flags usage
# go run ./cmds/exocomp/main.go tty architect --url="http://localhost:11434/api" --model="qwen3-coder:30b";Dual Licensed. AGPL3 for private usage. EULA for commercial usage available. For a commercial license, contact Cookie Engineer.
As you might have imagined, this is a not-so-serious project at this stage. Maybe it works, maybe it doesn't. Only the future will be able to tell whether the LLM hype of agentic coding/debugging environments was justified.