Your codebase. Your workflow.
An open-source coding agent for the terminal. Choose a supported model provider, connect your tools, and move between exploration, planning, and code changes.
Tools for the development loop
Explore the capabilities documented in the public repository, from code navigation to reusable workflows and agent coordination.
Multi-Provider
Choose among supported providers including xAI, Anthropic, OpenAI, Ollama, Groq, and DeepSeek, with retry and failover controls.
Codebase Tools
Search files, inspect code, make edits, and run shell commands from a terminal conversation. Review changes through diff and undo tools.
MCP Support
Configure MCP servers to connect external tools and resources through supported stdio, SSE, and WebSocket transports.
Reusable Skills
Load Markdown skills as slash commands for repeatable tasks, alongside built-in commands for models, sessions, planning, and tools.
Plan Mode
Explore a codebase in read-only planning mode before moving into execution. Configure tool permissions for the work you want to allow.
Smart Context
Context compression and a usage display help manage longer conversations. Save and resume sessions as work continues.
Agent Coordination
Coordinate sub-agents, task queues, and named teams, with worktree tools for separating changes across parallel tasks.
Image Support
Drag-and-drop images directly into the terminal. Smart paste collapses large payloads automatically to preserve context.
From a request to a reviewable change
An example of the intended workflow: inspect relevant files, plan edits, and review the result. This illustration is not a recorded run or a performance benchmark.
Start with the current project instructions.
AshlrCode runs on Bun. The source tree currently references sibling Ashlr packages, so review the repository and its dependencies before setting up your environment. Configure the supported provider and tool permissions you intend to use.
Hosted models require their provider configuration and may incur usage charges. Local models use an Ollama setup. Available features and context limits depend on the selected model and configuration.
Start coding with AshlrCode
Inspect the implementation, follow development, and contribute to the tools behind the workflow.