Roadmap

Shipped CLI workflows include interactive setup and installation, linting, skill and memory management, date-prefixed feature-document initialization, plugins, Telegram and Slack channels, durable task tracking, and the multi-agent management console. Remaining work focuses on consistent command ergonomics, clearer diagnostics, stronger non-interactive workflows, and reducing friction across these established surfaces rather than adding another broad command category.

Agent-specific hooks are partially shipped. AI DevKit can install Codex session-mapping hooks for more reliable session discovery and Claude hooks that forward channel-related tool activity and questions. A general hook automation system remains future work. That broader milestone would let teams define portable chain reactions, such as requesting review after implementation, instead of relying only on the integrations AI DevKit configures today.

Memory management is underway. AI DevKit supports storing, searching, and updating entries, while the first-party memory dashboard and the agent console's recent-memory pane provide browse-oriented views. Dedicated merge and delete workflows, richer curation, and deeper browsing controls remain future work so knowledge bases can stay clean as projects evolve.

Trust that your AI is remembering the right things. This feature will introduce a testing framework for your memory bank, allowing you to define "retrieval tests" that verify whether the correct coding standards and documentation are being prioritized when your AI agent searches for context.

Reduce costs and stay within token limits without losing context. Context Compaction will implement intelligent summarization algorithms and "context packaging" to compress large documentation or conversation histories into concise, token-efficient formats that can be easily passed to new agents or threads.

Launch a static website with landing page, comprehensive documentation, project vision, and development roadmap.

AI DevKit now supports a broad set of coding agents and environments, including Cursor, Claude Code, GitHub Copilot, Gemini CLI, Codex, opencode, and Antigravity, with additional environments still being tested. The goal remains the same: make AI DevKit the control plane between your preferred coding agents and your development workflow.

Build a local AI memory that learns from your coding habits, preferences, and team knowledge. This feature allows AI agents to access your skills, patterns, and project context directly from local, minimizing redundant prompts, speeding up workflows, and keeping your documents lightweight and focused.

Skill Management provides registries for discovering, installing, updating, and sharing reusable AI capabilities across projects. You can register project or global sources with `skill add-registry`, update every cached registry or one selected registry with `skill update [registry-id]`, and list, install, or remove skills in project and global agent locations.

The original agent-management milestone is complete. AI DevKit can detect and list active agents, inspect live and historical sessions, focus terminals, send messages to agents or groups, start and stop managed agents, rename agents, and run durable Claude print agents. The agent console adds previews, messaging, start/rename/kill controls, pinning, name filtering, a scrollable detail pane, recent memory, and channel controls. Future work is narrower: expand provider parity where local CLIs expose different capabilities, improve multi-agent orchestration ergonomics, and continue hardening terminal and session detection across platforms.