| FMRS | 69 / 100 · C | 75 / 100 · B |
|---|
| Reliability | 10 / 20 | 12 / 20 |
|---|
| Security and permissions | 15 / 20 | 16 / 20 |
|---|
| Maintenance | 13 / 20 | 14 / 20 |
|---|
| Documentation | 17 / 20 | 18 / 20 |
|---|
| Setup experience | 14 / 20 | 15 / 20 |
|---|
| Best for | - Users whose machine can carry local diffusion (a discrete GPU with >= 24 GB VRAM is ideal; Apple Silicon with >= 32 GB unified memory is OK for images)
- Users who want to automate a local ComfyUI workflow from Claude Code, Claude Desktop, Cursor or any MCP-speaking client
- Users who need to introspect live nodes, models and templates, custom nodes included
- Users running a local LLM and ComfyUI on the same GPU who need the read/free VRAM loop
| - Individuals or teams who prioritise privacy and offline operation and do not want data leaving the machine
- Users who share memory across multiple MCP clients such as Claude Code, Claude Desktop, Hermes Agent, OpenClaw, Cursor, Codex and Gemini CLI
- Scenarios that need encrypted storage, a hash-chained audit log and verifiable vault migration
- Users who want low-latency recall, a bundled embedding model and no extra LLM calls inside the memory layer
|
|---|
| Not for | - Users who want Comfy Cloud execution, cloud queue or cross-session cloud batches — this server has no path to Comfy Cloud, use https://cloud.comfy.org/mcp instead
- Machines with under 8 GB VRAM or a confirmed absence of a GPU (use partner nodes or the cloud MCP)
- Users who need lifecycle management (launch/stop/update/version-switch/install nodes/logs) on a remote box — those tools are local-only
- Users running the MCP on a remote or containerised host who need the browser sign-in callback to land on their machine
| - Users who need cloud multi-device sync or a hosted team memory service; moving between machines means locking the vault and copying the file
- Users who expect the server itself to run an LLM that extracts facts and decides what to remember; Compartment explicitly keeps no LLM inside and leaves that to the host model
- Users who cannot keep a passphrase safe or who need a recovery path if it is lost, since Compartment never generates a password, seed or recovery phrase
- Deployments that require network transports such as SSE or streamable-http; this server is stdio only and opens no ports
|
|---|
| Required permissions | - Runs the comfy command as a subprocess on your machine, carrying the server process's full environment
- Reads ComfyUI workspace files, including workflow JSON and input assets
- Writes generated outputs, by default into the ComfyUI workspace's output/ directory; fetch_outputs can copy them to a directory you name
- Writes/installs models and node packs on this machine (download_model, install_node)
- Reads, launches, stops or restarts the local ComfyUI process
- launch_comfyui with --listen exposes an unauthenticated ComfyUI to the network and requires explicit confirmation
- COMFY_API_KEY is used for partner-API nodes and can spend Comfy credits through those nodes
- COMFY_MCP_ASSUME_CONSENT can pre-authorize the install_node, update_all, version_switch, kill_untracked and network_exposure gates (spending is excluded)
| - Read and write the vault file under the user's home directory, by default ~/.compartment/memory.vault, and its sibling settings file
- Access the session directory holding the unlock credential (configurable with COMPARTMENT_SESSION_DIR), which also carries the shared embedding process's Unix socket
- Optionally use the macOS keychain (compartment unlock --keychain is an explicit opt-in, and it survives reboots)
- Write or merge an mcpServers entry into each MCP client's own configuration file, taking a byte-exact backup first
- Install a PostToolUse hook for Claude Code that captures memory files the agent writes
- Serve a read-only dashboard on 127.0.0.1 behind a one-time random URL token
|
|---|
| Risks and side effects | - Partner generation (partner_generate, workflows or templates containing partner-API nodes) really spends Comfy credits and is confirmed at call time; an agent host's 'always allow this tool' toggle is never treated as spend consent
- An arbitrary custom node can still call a paid service of its own — the confirm_spend gate only covers partner-API nodes comfy-cli recognizes, so check a workflow you did not build before running it
- launch_comfyui with --listen binds an unauthenticated ComfyUI to your network, exposing it to others
- Installing node packs, switching versions, updating ComfyUI and force-restarting a server it did not start modify your local environment; tools fail closed when the client does not show the confirmation prompt
- With a remote COMFYUI_URL configured, system_stats / free_memory still act on the local install, so you may measure and free the wrong machine; the remote ComfyUI must be reachable and unauthenticated on that network
- COMFY_MCP_ASSUME_CONSENT is a setting written in your client config; granting too much silently widens the risks above
- The server is beta, so behaviour may change
| - A leaked passphrase means the encrypted content can be decrypted; Compartment generates no recovery phrase and cannot recover a lost passphrase
- Enabling the memory_unlock tool places the passphrase in the model's context, which is why it is off by default
- If a 2FA keyfile is lost, for example a USB stick, the vault cannot be opened even with the passphrase
- `export --plaintext` writes the vault as unencrypted JSONL, so that file must be handled carefully
- Memory content can come from untrusted sources; recall wraps memories with a notice that they are stored data and content can be marked quarantined, but the host agent must still treat memory as data to limit prompt-injection risk
- The capture hook and client integration write into the user's own settings files, although the implementation backs up and merges rather than replacing
|
|---|
| Supported clients | Claude Code, Claude Desktop, Cursor | Claude Desktop, Claude Code, Hermes Agent, OpenClaw, Cursor, VS Code, Codex, Gemini CLI, Cline, Roo Code, Zed, OpenCode, LM Studio, AnythingLLM, BoltAI, Goose, Kiro |
|---|
| Tools | 30 | 14 |
|---|