Mind AI工作流

Requirements

• Bun 1.2+ (auto installed by the one-line installer if not present)

Installation

You can install Mind quickly and start using it right away. If you want agent setup later, jump to Agent setup.

One-line installer (recommended)

curl -fsSL https://raw.githubusercontent.com/GabrielMartinMoran/mind/main/scripts/install.sh | bash

This installs Mind to user-local paths (no sudo):

• app: ~/.local/share/mind
• launcher: ~/.local/bin/mind

Then run:

mind help

The installer creates a launcher that runs the installed src/mind.ts entry point. It doesn't configure every supported agent automatically. To refresh existing mind-managed agent integrations after installation, run mind setup refresh.

Install from source

git clone https://github.com/GabrielMartinMoran/mind.git
cd mind
bun install

Agent setup

Mind supports multiple agent integrations, but they are not all equally mature. Run mind setup without an agent name to see the current capability matrix, then configure the specific agent you want.

mind setup              # show capability matrix first
mind setup claude-code
mind setup opencode
mind setup cursor
mind setup codex
mind setup windsurf
mind setup gemini-cli
mind setup vscode
mind setup antigravity
mind setup refresh      # refresh detected mind-managed integrations

mind setup refresh is conservative by default. It refreshes only detected integrations, where detection means mind finds an existing mind-owned signal, such as an MCP config, managed protocol block, hook/plugin/script, or installed mind-management skill. Use --dry-run to preview changes, --all to refresh all supported agents, or --agent <name> to target one agent.

Post-install configuration

Most configuration is optional. You only need extra setup if you want to change paths, ports, or enable semantic search.

1. Create your .env file:

cp .env.example .env

Mind ships with a .env.example that contains all configurable options. The installer will also do this automatically if .env doesn't exist.

2. Configure environment variables (optional):

Edit .env to customize your setup:

RAG / semantic search setup:

To enable AI-powered semantic search:

```bash

Quick start

This is the fastest way to see what Mind feels like in practice.

```bash

Basic CLI example

mind create projects/mind "Mind project memory"
mind add projects/mind architecture "CLI uses command registry + atomic command modules"
mind search architecture

For the full command list:

mind help

Usage

In .env:

MIND_RAG=true OPENAI_API_KEY=sk-your-key-here ```

When enabled, memories are embedded using OpenAI's text-embedding-3-small model and combined with full-text search for hybrid retrieval.

mind autosync config

version: 1

Agent integration details

Use the commands in Installation to run setup. This section explains how to read the capability matrix and what each integration actually configures.

mind setup (without agent) now prints a capability matrix per integration using a 3-level model:

• L1: MCP transport wiring
• L2: instruction/protocol injection
• L3: hooks/session/compaction automation

Each level is explicitly marked as supported, unsupported, or unverified with evidence/fallback notes. If a capability is not implemented, setup output is explicit (no silent skip).

Agent status matrix

Status labels used here:

• Complete: L1/L2/L3 are all supported
• Partial: at least one level is unsupported or unverified
• Experimental: declaration exists but integration is explicitly unstable/unverified
• Roadmap: planned declaration only, no adapter wiring

Rollout policy:

• Wave 1 priority agents: OpenCode, Claude Code, Gemini CLI, Cursor
• Claude Code now includes managed L2 protocol injection by writing ~/.claude/instructions/mind-memory-protocol.md and maintaining a managed block in ~/.claude/CLAUDE.md
• Claude Code L3 hooks automation is opt-in and non-blocking (default off). Enable with MIND_SETUP_CLAUDE_ENABLE_HOOKS=true before running setup.
• Cursor L2 remains intentionally unverified (no verified global user-rules injection path)
• Cursor L3 is implemented with global hooks wiring (~/.cursor/hooks.json + managed hook script)
• Existing integrations outside Wave 1 remain wired in the same capability model with explicit status
• Antigravity is now a supported agent with L1 MCP wiring and skill installation at ~/.gemini/antigravity/

mind setup opencode is idempotent and non-destructive:

• prefers ~/.config/opencode/opencode.jsonc over opencode.json when both exist, and never creates opencode.json if opencode.jsonc is already present. New files are created as opencode.jsonc (the JSONC superset of JSON that OpenCode accepts natively).
• preserves unknown keys already present in the chosen config file
• configures mcp.mind as local command transport (type: "local", command: ["<path-to-mind>", "mcp"])
• writes/refreshes ~/.config/opencode/instructions/mind-memory-protocol.md
• ensures that instruction file is present in OpenCode's instructions list
• configures prudent L3 session/compaction automation by default and non-blocking, writing ~/.config/opencode/plugins/mind-automation.js during setup

Setup safety contract: all setup flows that touch a user-owned JSON/JSONC config file (opencode, claude fallback, cursor, windsurf, gemini-cli, vscode, antigravity) now go through src/setup/safe-config.ts. Existing files are parsed strictly — a parse failure aborts the run with a clear error rather than overwriting the file with an empty default. Content-changing writes are preceded by a timestamped sibling backup (.bak.YYYYMMDDTHHMMSSmmmZ) and the new content is staged in a sibling .tmp file and renamed into place so a crash mid-write cannot leave a partial file. The same contract applies to mind setup refresh.

mind setup codex keeps setup idempotent and writes local MCP command args in ~/.codex/config.toml:

• [mcp_servers.mind]
• command = "<path-to-mind>"
• args = ["mcp"]

It also non-destructively upserts a managed Memory Protocol block in ~/.codex/AGENTS.md.

mind setup cursor keeps existing MCP setup behavior and now configures global L3 continuity hooks:

• writes/updates ~/.cursor/hooks.json with managed idempotent entries for sessionStart, preCompact, and stop
• writes/refreshes executable hook script ~/.cursor/hooks/mind-session-continuity.sh
• preserves existing hook config keys/entries

Check server process status:

mind server-status