MiniCode Roadmap

MiniCode Roadmap

MiniCode already has a usable lightweight terminal coding workflow, but there is still a visible gap between the current `main` branch and a more complete Claude Code-like runtime.

This roadmap highlights the most valuable missing capabilities and the order in which they should ideally be improved.

Pull requests are welcome, especially when they align with the contribution guidelines and keep the project lightweight.

P0

1. Model-aware context management

**Status: in progress.**

This is the most important missing runtime capability.

It includes:

• model-aware context window configuration
• provider-reported usage accounting
• context usage display in the TUI
• automatic context compaction for long conversations

This work matters because long-session stability depends on it. It is also one of the most important design areas where MiniCode still trails a more complete Claude Code-style runtime.

2. API retry and backoff

**Status: implemented.** The main Anthropic adapter retries on 429 and 5xx with exponential backoff and honors `Retry-After` when present. The companion [Python](./external/MiniCode-Python/) and [Rust](./external/MiniCode-rs/) ports follow the same behavior.

Possible follow-ups include a more uniform configurable retry policy, richer observability, or applying the same policy consistently across additional provider code paths.

3. Session persistence and resume

MiniCode should be able to save and resume sessions reliably.

This includes:

• autosave
• manual resume
• basic session recovery

This is important for real-world usage and longer task execution.

4. Multi-language implementation branches

**Status: Python and Rust companion implementations are available; Go remains exploratory.**

Another important direction is to explore parallel implementations of MiniCode in other languages, especially:

• Python (companion repo available)
• Go (no formal companion implementation yet)
• Rust (companion repo available)

This is particularly valuable for the learning side of the project.

The goal is not to fragment the main codebase immediately. The goal is to encourage language-specific branches or companion implementations that preserve the same core ideas:

• lightweight architecture
• Claude Code-aligned design direction
• readable agent loop and tool model
• educational value for contributors studying different ecosystems

If you are interested in maintaining or extending the Python or Rust variants, or starting a Go direction, contributions and direct collaboration are welcome.

P1

5. Layered memory loading

MiniCode should support a lightweight memory hierarchy similar in spirit to Claude Code's layered project context.

This may include:

• global memory
• project memory
• nested/project-local memory
• simple include support where appropriate

6. Stronger provider abstraction

MiniCode currently works well with Anthropic-style APIs and some compatible providers, but th