← Back to products
Autonomous agents

MantisCoder

An autonomous coding agent that understands your whole repository.

A desktop coding agent with a tree-sitter code graph, an agentic task loop, sub-agents, and native control of the browser, terminal and git.

mantiscoder

MantisCoder indexes a repository with tree-sitter into a dependency and call graph, then plans and executes coding tasks against that graph — not against a context window. Every task follows explicit rules for code changes, file operations, refactoring, testing and completion.

Sub-agents take on bounded work, the browser control module verifies UI changes, and the terminal and git modules run builds, tests and commits. System prompts and task rules are plain text files you can edit.

Highlights
  • Editable system prompts: coding assistant, browser control, terminal control, git operations
  • Task rules for code changes, file operations, refactoring, testing and completion
  • Interactive HTML graphs of the repository
  • Bug manager, test lab, kanban board and task manager surfaces
  • Built-in browser, connector services and MCP connections
  • Local SQLite project memory
Key capabilities
01

Code graph

Tree-sitter analysis builds dependency, call and radial graphs of the whole project.

02

Agentic loop

Plan → act → evaluate with explicit task evaluation rules and completion criteria.

03

Sub-agents

Bounded sub-tasks delegated with their own instructions and scope.

04

Browser, terminal, git

Native control modules verify UI, run builds and tests, and commit.

In the product
Technical profile
Analysis
tree-sitter, multi-language
Memory
Local SQLite
Control
Browser, terminal, git
Platform
Desktop, Go + native shell
Use cases
Large legacy codebasesRefactoring campaignsTest generationInternal developer tooling
YAPZEK®

Ready to put AI on your own infrastructure?

Tell us about your environment, your data boundary and your latency target. We will design the pipeline with you and bring it live within days, not months.