Waggle MCP server
{
"mcpServers": {
"abhigyan-shekhar-waggle": {
"command": "uvx",
"args": [
"waggle-mcp"
]
}
}
}{
"mcpServers": {
"abhigyan-shekhar-waggle": {
"command": "uvx",
"args": [
"waggle-mcp"
]
}
}
}{
"mcpServers": {
"abhigyan-shekhar-waggle": {
"command": "uvx",
"args": [
"waggle-mcp"
]
}
}
}{
"servers": {
"abhigyan-shekhar-waggle": {
"command": "uvx",
"args": [
"waggle-mcp"
]
}
}
}{
"mcpServers": {
"abhigyan-shekhar-waggle": {
"command": "uvx",
"args": [
"waggle-mcp"
]
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall uvx waggle-mcpversion 0.1.8
What it does
Waggle is a local-first memory layer for AI agents that stores project knowledge as a graph, preserving decisions, context, and reasoning across conversations. It works with MCP clients and provides tools for capturing, connecting, and recalling project memory without repeated context pasting.
- Recall project decisions and requirements across multiple conversation sessions without re-explaining context.
- Track the reasoning and supporting evidence behind decisions, including contradictions and how they changed over time.
- Export and import project memory graphs (`.abhi` files) to move memory between different MCP clients and workflows.
Tools this server exposes
| Tool | Does |
|---|---|
| prime_context | Load project context at the start of a session. |
| query_graph | Retrieve relevant history and memories from the project graph. |
| observe_conversation | Record durable outcomes and decisions from the current conversation into memory. |
| build_context | Assemble a compact context pack for a specific task. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.