Persistent graph-backed conversational memory for AI agents.
Waggle MCP server is a hosted integration for AI assistants that speak the Model Context Protocol. Persistent graph-backed conversational memory for AI agents.
Persistent memory that remembers decisions, reasons, and contradictions across sessions.
Once Waggle is connected, these are the calls the assistant has available:
observe_conversation — After any turn containing a decision, preference, constraint, correction, or project fact. Persists the verbatim turn first, then extracts graphquery_graph — Before answering questions that may depend on prior context. Hybrid retrieval (graph + verbatim transcript) by default. Supports as_of forprime_context — At the start of a new session to hydrate context from the most relevant scoped memoriesgraph_diff — When the user asks what changed recentlyupdate_node — Update an existing node's content, label, or tagsdelete_node — Delete a node and all its edgesdecompose_and_store — Break long content into atomic nodes and infer edges automaticallydedup_candidates — Return near-duplicate node pairs above a similarity threshold for human reviewcanonicalize_node — Merge multiple nodes into one canonical node. Repoints all edges, collects aliases. Idempotentedge_quality_report — Audit edge quality — counts, average confidence per type, top/bottom confidence edgesdebug_retrieval — Diagnose retrieval ranking for a query — embedding scores, window routing, tiered vs flat comparisonCodex — Waggle is available for Codex as a self-hosted local MCP plugin. No hosted backend, Apple Developer ID, or Windows Authenticode certificate isYou will need 6 environment variables: WAGGLE_TRANSPORT, WAGGLE_BACKEND, WAGGLE_DB_PATH, WAGGLE_DEFAULT_TENANT_ID, WAGGLE_MODEL, WAGGLE_STARTUP_MODE. Keep credentials in your client's env block or a secrets manager rather than in a file you might commit.
Being a remote server, there is no local install. You register the endpoint with your client, authorise it once, and the tools appear.
Among the developer tooling options, the useful question is rarely "what can it do" but "what does it cost you to run" — permissions, credentials, and how much of your context its toolset consumes. Waggle's toolset — observe_conversation, query_graph, prime_context and 11 more — is a fair guide to whether it matches your workflow. It is maintained by Abhigyan-Shekhar; worth a glance at recent repository activity before you build anything load-bearing on it.
SyncDev reviews every entry in this directory against the project's own documentation before publishing, and revisits them as servers change.
| Tool | What it does |
|---|---|
| observe_conversation | After any turn containing a decision, preference, constraint, correction, or project fact. Persists the verbatim turn first, then extracts graph nodes. Returns turn_id, verbatim_stored, nodes_extracted, edges_inferred. |
| query_graph | Before answering questions that may depend on prior context. Hybrid retrieval (graph + verbatim transcript) by default. Supports as_of for point-in-time queries. |
| prime_context | At the start of a new session to hydrate context from the most relevant scoped memories. |
| graph_diff | When the user asks what changed recently. |
| update_node | Update an existing node's content, label, or tags. |
| delete_node | Delete a node and all its edges. |
| decompose_and_store | Break long content into atomic nodes and infer edges automatically. |
| dedup_candidates | Return near-duplicate node pairs above a similarity threshold for human review. |
| canonicalize_node | Merge multiple nodes into one canonical node. Repoints all edges, collects aliases. Idempotent. |
| edge_quality_report | Audit edge quality — counts, average confidence per type, top/bottom confidence edges. |
| debug_retrieval | Diagnose retrieval ranking for a query — embedding scores, window routing, tiered vs flat comparison. |
| Codex | Waggle is available for Codex as a self-hosted local MCP plugin. No hosted backend, Apple Developer ID, or Windows Authenticode certificate is required: users download the marketplace zip from GitHub Releases, add it to |
| Cursor | Cursor Settings → Features → MCP Servers → + Add - Command: waggle-mcp - Args: serve - Env vars: same keys as the JSON block above. |
| Antigravity | The AI agent reads ~/.gemini/antigravity/mcp_config.json (macOS/Linux) or %USERPROFILE%\.gemini\antigravity\mcp_config.json (Windows). Add the waggle block there. The VS Code extension panel reads a different file — addi |
{
"mcpServers": {
"waggle": {
"command": "waggle-mcp",
"args": ["serve"],
"env": {
"WAGGLE_TRANSPORT": "stdio",
"WAGGLE_BACKEND": "sqlite",
"WAGGLE_DB_PATH": "~/.waggle/waggle.db",
"WAGGLE_DEFAULT_TENANT_ID": "local-default",
"WAGGLE_MODEL": "all-MiniLM-L6-v2",
"WAGGLE_STARTUP_MODE": "normal"
}
}
}
}Configuration as documented by the project. Restart the client after saving.
| Variable | Description | Required |
|---|---|---|
| WAGGLE_TRANSPORT | Configuration value read at startup. | Optional |
| WAGGLE_BACKEND | Configuration value read at startup. | Optional |
| WAGGLE_DB_PATH | Filesystem location the server is allowed to use. | Optional |
| WAGGLE_DEFAULT_TENANT_ID | Configuration value read at startup. | Optional |
| WAGGLE_MODEL | Configuration value read at startup. | Optional |
| WAGGLE_STARTUP_MODE | Configuration value read at startup. | Optional |
Kill hallucinated APIs — version-accurate, up-to-date library documentation injected straight into context.
Microsoft's official browser automation server — drive a real browser through the accessibility tree, no screenshots needed.
GitHub's official server — repos, issues, pull requests, Actions and code security, straight from your assistant.
Issue tracking at the speed of conversation — Linear's official hosted server with OAuth and zero install.
Local repository surgery — status, diffs, commits, branches and history for any repo on disk.
Timezone sanity for AI — current time anywhere and correct conversions, without the model doing date math.