AnythingLLM Control Plane MCP Server
This MCP server lets you administer AnythingLLM workspaces and keep their vector stores in sync with local Markdown documentation via an ETL daemon.
list_workspaces: List all AnythingLLM workspaces with slugs, IDs, and vector counts (read-only).
create_workspace: Create a new isolated workspace with custom settings.
delete_workspace: Permanently delete a workspace and purge its vector storage (destructive).
get_system_env: Retrieve system environment, vector DB provider, and storage diagnostics.
get_vector_count: Get vector counts system-wide or for a specific workspace slug.
anythingllm-sync daemon: Scan project directories for Markdown files, deduplicate via SHA-256, upload new/modified docs to matching workspaces, and remove vector entries when files are deleted locally.
Operational flexibility: Run as a stdio MCP server, configure via environment variables or CLI flags, run one-shot syncs or continuous intervals.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@AnythingLLM Control Plane MCP Server List all my workspaces."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
module_type: mcp-server status: active protocol: mcp primary_capability: AnythingLLM administration, workspace lifecycle, and automated ETL vector sync requires: AnythingLLM instance works_with: AnythingLLM, Antigravity Agent Ecosystem, MCP Clients last_verified: 2026-09-05
AnythingLLM Control Plane & ETL Sync MCP Server ๐ง
High-performance Go-based Model Context Protocol (MCP) server and automated ETL sync daemon for AnythingLLM instances (TheNovaNodes/anythingllm-mcp-control). Provides autonomous AI agents with administrative workspace management and keeps semantic vector stores in continuous synchronization with filesystem Markdown documentation.
Related MCP server: anythingllm-mcp
๐ฆ Core Binaries & Capabilities
The project builds two specialized Go binaries:
anythingllm-controlโ Stdio-based MCP server exposing an administrative interface for managing AnythingLLM workspaces, vector indices, and system diagnostics.anythingllm-syncโ Autonomous ETL synchronization CLI and daemon that monitors local project directories, deduplicates content via SHA-256, and handles tombstone lifecycle tracking in pure Go SQLite.
๐ ๏ธ Exposed MCP Tools
The anythingllm-control MCP server exposes 5 administrative tools:
list_workspaces
Lists all active workspaces in AnythingLLM with their slugs, IDs, and vector counts.
Arguments: None.
Side Effects: None (read-only).create_workspace
Creates a new isolated workspace with custom settings and vector indexing.
Arguments:name(string, required).
Side Effects: Creates a new workspace in the AnythingLLM database.delete_workspace
Permanently removes a workspace and purges its associated vector storage.
Arguments:slug(string, required).
Side Effects: Destructive permanent deletion of the workspace.get_system_env
Retrieves system environment settings, vector DB provider, and storage diagnostics dump.
Arguments: None.
Side Effects: None (read-only).get_vector_count
Retrieves total vector counts system-wide or for a designated workspace slug.
Arguments:slug(string, optional).
Side Effects: None (read-only).
๐ Quick Start & Building
Prerequisites
Go 1.25 or higher
GCC or Clang (optional, pure Go SQLite is used)
Build Binaries
git clone https://github.com/TheNovaNodes/anythingllm-mcp-control.git
cd anythingllm-mcp-control
make buildCompiled binaries will be created in ./bin/:
bin/anythingllm-controlbin/anythingllm-sync
Install System-wide
sudo cp bin/anythingllm-control /usr/local/bin/
sudo cp bin/anythingllm-sync /usr/local/bin/Health Check (stdio smoke test)
anythingllm-control < /dev/nullThe server will log its initialization status and exit cleanly upon receiving EOF.
โ๏ธ Configuration & Environment Variables
The server and sync daemon read configuration from environment variables or command-line flags:
ANYTHINGLLM_BASE_URL(orALM_BASE)
Base URL of the AnythingLLM REST API.
Default:http://127.0.0.1:3002/api/v1ANYTHINGLLM_API_KEY(orMG_API_KEY)
Bearer API key for AnythingLLM authentication.
๐ Automated ETL Sync Daemon (anythingllm-sync)
anythingllm-sync scans repositories for Markdown documentation, calculates SHA-256 checksums, and uploads new or modified documents into matching AnythingLLM workspaces. When a tracked file is deleted locally, it records a tombstone and removes the document from the vector index.
CLI Flags
-projects(string)
Root directory containing projects or agent offices to scan.
Default:/root/projects-state-dir(string)
Directory storing the local SQLite state ledger (sync_state.db).
Default:/root/projects/TheNovaNodes/ops/shared/anythingllm-sync-alm-base(string)
AnythingLLM API base URL.
Default:http://127.0.0.1:3002/api/v1-api-key(string)
AnythingLLM Bearer API key.-once(bool)
Execute a single synchronization pass and exit immediately.-interval(duration)
Continuous daemon run interval (e.g.15m,1h). Default0runs once.-timeout(duration)
HTTP request timeout per operation.
Default:30s
Example Usage
# One-shot scan and index
anythingllm-sync -projects /root/projects -api-key "$ANYTHINGLLM_API_KEY" -once
# Continuous daemon running every 30 minutes
anythingllm-sync -projects /root/projects -api-key "$ANYTHINGLLM_API_KEY" -interval 30m๐ MCP Client Configuration
Add to your MCP client configuration (e.g., Claude Desktop, Antigravity, or mcp-router):
{
"mcpServers": {
"anythingllm-control": {
"command": "/usr/local/bin/anythingllm-control",
"args": [],
"env": {
"ANYTHINGLLM_BASE_URL": "http://127.0.0.1:3002/api/v1",
"ANYTHINGLLM_API_KEY": "YOUR_API_KEY_HERE"
}
}
}
}๐งช Testing & Code Quality
Run tests with data race detection and coverage:
# Run unit tests
make test
# Run tests with HTML/func coverage report
make coverage
# Run Go static analysis
make lint๐ Security Boundaries
Strict Access Control:
anythingllm-controlcontains destructive capabilities (delete_workspace). In production setups, it must only be exposed to trusted administrative agents or partitioned viamcp-routerACLs.Prompt Injection Defense: Input strings for workspace names and slugs are sanitized to prevent path traversal and API injection.
Zero-Secret Leakage: API keys are injected via environment variables and never logged or serialized in error outputs.
๐ License
MIT License โ see LICENSE for full details.
Available Tools
5 toolscreate_workspaceB
Create a new workspace in AnythingLLM.
Args: name: Name for the new workspace.
Returns: JSON string with created workspace details or error message.
Inputs: name (required). Returns: result
delete_workspaceB
Delete a workspace in AnythingLLM by slug.
Args: slug: The workspace slug to delete.
Returns: JSON string or status message confirming deletion or reporting error.
Inputs: slug (required). Returns: result
get_system_envA
Retrieve system environment and configuration dump from AnythingLLM Admin API.
Returns: JSON string containing system environment variables and configuration.
Inputs: no named parameters. Returns: result
get_vector_countA
Get vector count for AnythingLLM (system-wide or workspace-specific).
Args: slug: Optional workspace slug. If provided, returns vector count for that workspace.
Returns: JSON string or status message with vector count stats.
Inputs: slug. Returns: result
list_workspacesA
List all workspaces in AnythingLLM.
Returns: JSON string containing the list of workspaces or error status.
Inputs: no named parameters. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelogโฆ
TDQS
Scored across 5 tools
Each tool targets a distinct resource and action: workspace CRUD operations (list/create/delete) are clearly separated from system environment and vector count queries. No two tools have overlapping purposes, making selection unambiguous.
All tool names follow a consistent verb_noun pattern in snake_case: list_workspaces, create_workspace, delete_workspace, get_system_env, get_vector_count. The verbs and nouns are predictable and match the action/resource pair.
Five tools is well-scoped for a control plane server, covering core workspace management and system information without unnecessary bloat. The count feels appropriate for the apparent domain.
The workspace lifecycle is incomplete: create, list, and delete exist, but update is missing. System environment is read-only, and there is no way to modify configuration. These gaps are notable for a control plane server, though the core viewing/deleting operations are present.
Maintenance
Related MCP Connectors
Your org's AI agents, tasks, runs, search, and brain files as MCP tools and resources.
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
Build and run AI workflows, agents, data tables and 1000+ API integrations from any MCP client.
Manage SRG+ hubs, channels, content, assets, users, and workspaces from any MCP-aware AI agent.
Related MCP Servers
- AlicenseBqualityDmaintenanceEnables seamless integration with AnythingLLM instances, providing complete workspace management, chat operations, document handling, user administration, and AI agent configuration through natural language.3850 npm16MIT
- AlicenseCqualityDmaintenanceMCP server for AnythingLLM that enables AI document chat platform interaction with tools for workspaces, chat, documents, threads, and system operations.17MIT
- FlicenseNot gradedqualityDmaintenanceEnables AI agents and users to manage workspace files, monitor system metrics, take persistent notes, and retrieve weather data via MCP tools and resources.-
- AlicenseAqualityCmaintenanceEnables MCP-compatible clients to interact with AnythingLLM, providing tools for workspace management, chat and thread operations, document operations, vector search, and system inspection.346MIT