IDE Plugins & Configuration

Connect Awareness to your IDE agent through MCP. This page covers the public setup path for Cursor, Claude Code, Windsurf, Cline, VS Code GitHub Copilot, Codex, and OpenCode.

Universal MCP Config

Most MCP-capable IDEs use the same JSON shape. Replace the placeholders with your values:

{
  "mcpServers": {
    "awareness-memory": {
      "type": "http",
      "url": "https://awareness.market/mcp",
      "headers": {
        "Authorization": "Bearer aw_your-api-key",
        "X-Awareness-Memory-Id": "your-memory-id",
        "X-Awareness-Agent-Role": "builder_agent"
      }
    }
  }
}

If your IDE supports multiple MCP servers, you can register one server per role.

Claude Code

Option A: One-click install prompt

Awareness provides a one-click setup prompt in the Claude Code page inside the product UI. Open the page, choose a Memory, and copy the generated prompt. It will:

1. Read your existing .mcp.json (project root)
2. Merge the awareness-memory MCP server into mcpServers
3. Ask for your API key and Memory ID if they are missing
4. Verify the connection with claude mcp list
5. Test the integration with awareness_init

Option B: Manual MCP config

Config file: .mcp.json (project root)

{
  "mcpServers": {
    "awareness-memory": {
      "type": "sse",
      "url": "https://awareness.market/mcp",
      "headers": {
        "Authorization": "Bearer aw_your-api-key",
        "X-Awareness-Memory-Id": "your-memory-id",
        "X-Awareness-Agent-Role": "builder_agent"
      }
    }
  }
}

Optional rule file: CLAUDE.md in the project root. This is the strongest way to keep the workflow visible even when the conversation gets compressed.

Claude Code plugin

If you want a higher-level Claude Code workflow on top of MCP, install the plugin:

/plugin marketplace add edwin-hao-ai/Awareness-SDK
/plugin install awareness-memory@awareness

The plugin exposes these slash commands:

Expected behavior

• awareness_init runs at the start of each session. If _setup_action is returned, follow it to write rules.
• __awareness_workflow__ is callable — use it when unsure what to do next
• Every meaningful edit, decision, and bug fix should be persisted via awareness_record
• If a write returns _extraction_instruction, your client LLM MUST process it and submit the result with awareness_record(insights=<result>) — this is how knowledge cards are created

Cursor

Config file: .cursor/mcp.json

{
  "mcpServers": {
    "awareness-memory": {
      "url": "https://awareness.market/mcp",
      "headers": {
        "Authorization": "Bearer aw_your-api-key",
        "X-Awareness-Memory-Id": "your-memory-id",
        "X-Awareness-Agent-Role": "builder_agent"
      }
    }
  }
}

Optional managed rule file: .cursor/rules/awareness.mdc

Behavior:

• Cursor calls awareness_init when a Composer or chat session starts
• Use awareness_lookup for fast structured context
• Persist work with awareness_record after meaningful edits

Windsurf

Config file: ~/.codeium/windsurf/mcp_config.json

Use the same universal MCP config.

Optional rule file: .windsurfrules

Behavior:

• Call awareness_init at the start of each Cascade flow
• Use awareness_lookup(type="tasks") to resume work
• Call awareness_record after important steps

Cline

Config file: Cline MCP settings UI

Paste the universal MCP config into the MCP server configuration screen.

Optional rule file: .clinerules

Behavior:

• Call awareness_init when a new task starts
• Use awareness_recall before implementing
• Record edits, shell commands, and decisions with awareness_record

VS Code GitHub Copilot

Config file: .vscode/mcp.json

Use the same universal MCP config.

Optional rule file: .github/copilot-instructions.md

Behavior:

• Copilot Chat or Copilot agent should call awareness_init first
• Use awareness_lookup for fast task and knowledge reads
• Write important changes back with awareness_record

Antigravity

Use the same Universal MCP Config if your Antigravity workspace exposes MCP settings or custom tool configuration.

Recommended path:

1. Paste the universal MCP config into the Antigravity workspace tool settings
2. Keep the Awareness workflow summary in the repo rule file
3. Run awareness_init at the start of each task or agent session
4. Persist meaningful work with awareness_record

If your current Antigravity build does not expose MCP yet, use the same fallback as Codex below: keep the workflow in your repo instructions and use the SDK / HTTP path for explicit writes.

Augment

Augment works well with the same public setup path as Cursor and Windsurf:

1. If MCP configuration is available, paste the universal MCP config into the Augment agent settings
2. Keep the Awareness workflow visible in the repo rule file or project instructions
3. Start sessions with awareness_init
4. Use awareness_lookup for fast structured reads and awareness_record for write-back

This is the recommended public entry point until a deeper Augment-specific installer is published.

Devin

Devin is usually best connected at the repo or workspace level instead of as an IDE-local MCP install.

Recommended public path:

1. Put the Awareness workflow summary in AGENTS.md or your repo instructions
2. If the Devin environment exposes MCP settings, register the universal MCP config there
3. Otherwise, use the Awareness SDK / HTTP API from your workspace bootstrap scripts or automation steps
4. Record task handoff notes, PR context, and fix summaries back into the same memory

This keeps long-running cloud workspaces, PR review loops, and follow-up tasks attached to one shared memory even when native MCP is not available.

Codex (OpenAI)

Codex does not rely on MCP in the same way as Claude Code or Cursor. Use the rule-file approach:

1. Sync the Awareness managed policy into AGENTS.md
2. Use the REST/SDK path for explicit writes if MCP is not available
3. Keep the same workflow: initialize context, check recall, record meaningful progress

Kiro (Amazon)

Config file: .kiro/settings/mcp.json

Use the same universal MCP config.

Optional rule file: .kiro/steering/awareness.md

Behavior:

• Call awareness_init at the start of every new Kiro session
• After each code edit or spec update, call awareness_record
• MCP server is configured in .kiro/settings/mcp.json

Trae (ByteDance)

Config file: .trae/mcp.json

Use the same universal MCP config.

Optional rule file: .trae/rules/awareness.md

Behavior:

• Call awareness_init at the start of every new chat session
• After each code edit, call awareness_record
• MCP server is configured in .trae/mcp.json

Zed

Config file: ~/.config/zed/settings.json (under context_servers)

Note: Zed uses context_servers instead of mcpServers. Add the Awareness server entry to the context_servers section in your Zed settings.

Optional rule file: .rules

Behavior:

• Call awareness_init at the start of every new assistant panel session
• After each code edit, call awareness_record

JetBrains (Junie)

Config file: .junie/mcp/mcp.json

Use the same universal MCP config. You can also configure MCP via Settings > Tools > MCP Server in your JetBrains IDE.

Optional rule file: .junie/guidelines.md

Behavior:

• Call awareness_init when Junie starts a new task
• After each code edit or refactoring, call awareness_record

OpenCode

Config file: opencode.json

{
  "mcp": {
    "awareness-memory": {
      "url": "https://awareness.market/mcp",
      "headers": {
        "Authorization": "Bearer aw_your-api-key",
        "X-Awareness-Memory-Id": "your-memory-id",
        "X-Awareness-Agent-Role": "builder_agent"
      }
    }
  }
}

Multi-role Setup

If one memory contains multiple agent roles, register separate servers:

{
  "mcpServers": {
    "awareness-dev": {
      "url": "https://awareness.market/mcp",
      "headers": {
        "Authorization": "Bearer aw_your-api-key",
        "X-Awareness-Memory-Id": "your-memory-id",
        "X-Awareness-Agent-Role": "developer"
      }
    },
    "awareness-reviewer": {
      "url": "https://awareness.market/mcp",
      "headers": {
        "Authorization": "Bearer aw_your-api-key",
        "X-Awareness-Memory-Id": "your-memory-id",
        "X-Awareness-Agent-Role": "reviewer"
      }
    }
  }
}

Each role gets the same memory container with its own scoped view.

Troubleshooting