Roadmap

The project should grow in small, useful layers. The priority is to stay simple for CLI users while exposing deeper capabilities for editor, API, and automation workflows.

Now

• Installable CLI with agno-deep-agent and agdeep.
• Python API with create_deep_agent.
• Agno provider:model strings, including OpenAI, Anthropic, Google, Groq, local Ollama, and Ollama Cloud.
• SQLite sessions, memory, and learning.
• Local skills.
• GitHub Pages documentation.
• ACP stdio server with agdeep acp.
• ACP editor integration guide for Zed, VS Code-compatible clients, JetBrains, and Neovim.

Next

• ACP client config snippets per editor/plugin ecosystem.
• CLI profiles for reusable local configuration.
• Safer permission presets for shell and filesystem tools.
• Better session resume UX.
• More examples for coding, docs, and repository review.

Planned

ACP Adapter Improvements

Expose richer Agno Deep Agent progress over ACP for editor and IDE clients.

Current command:

agdeep acp --model ollama:gemma4:e4b --workspace .

Implemented:

• stdio server entrypoint;
• editor session to Agno session mapping;
• basic plan, message, session metadata, and tool updates.

Planned work:

• richer member and task progress events;
• stronger cancellation for long model/tool calls;
• editor-friendly file and shell events;
• Zed, VS Code, JetBrains, and Neovim examples.

MCP Tool Loading

Attach MCP tools from a user config file without confusing MCP with ACP.

AgentOS Example

Add an Agno-native deployment example for users who want an HTTP/API surface.

Human Approval Policy

Add explicit approval rules for high-risk tool calls and destructive changes.

Later

• Container or remote sandbox backend.
• Rich local event viewer.
• Packaged skill examples.
• CI for docs and package publishing.