Terminal-native coding agent

Proxify Code

A developer agent for long-running code work: plan, edit, run commands, keep project memory, and move through implementation without leaving the terminal.

proxify-code
$ proxify code --goal "ship the billing fix"
plan

Map the repository and split the work into verifiable tasks.

edit

Patch the codebase with controlled file and command permissions.

verify

Run focused checks, summarize risks, and prepare the handoff.

memory

Persist project notes, checkpoints, and task progress between sessions.

The useful parts are agentic, not cosmetic.

OpenCode foundation

Provider support, TUI, MCP, LSP, plugins, shell tools and coding-agent primitives are the base layer.

Long-horizon work

Project memory, checkpoints, task trees, workflows and subagents shape longer coding sessions.

Proxify layer

Brand, onboarding, billing, hosted models and product UX stay aligned with the Proxify AI workspace.

Roadmap

A practical build path

01 Rebrand cleanly

Use Proxify names, packages and config paths while preserving MIT attribution.

02 Keep upstream compatible

Track OpenCode core changes separately from Proxify-specific agent features.

03 Ship as a product

Connect login, plans, hosted models and support to the existing Proxify AI backend.

Attribution

Proxify Code is planned as a Proxify product layer based on the MIT-licensed OpenCode project, with selected long-horizon agent ideas inspired by MiMo-Code. OpenCode and MiMo names remain owned by their respective projects.