Manages your plug-in server list.
Satcheli is where you organize the Model Context Protocol (MCP) servers AiMANAC and your other AI tools talk to. One purchase, $1, separate App Store listing. It pairs with AiMANAC — same sign-in, same trust mesh — but keeps the registry in its own on-device vault so a future read-only sibling could browse servers without ever seeing your sign-in material.
AiMANAC alone covers the “paste one server and go” case. The moment you juggle more than one — a local helper, a catalog integration, and your own customer backend — you want a dedicated organizer. Satcheli is that rack: add servers once, tune who may call them, and keep background health checks off your main console.
Five tabs:
- Servers — searchable list. Tap a row to edit the four control axes, per-tool allow/deny, and optional background “butler” polling.
- Absorb — paste a manifest, import a file from Files, or pull starter entries from public catalogs (Smithery and Anthropic’s directory today; more can land later).
- XZA Suite — granular switches for the twelve Everplay helper tools instead of the single master toggle AiMANAC shows by default.
- Butlers — glanceable dashboard for scheduled polls: last run, next run, and errors without opening each server card.
- Settings — defaults for new servers, optional catalog API keys, suite reset, and a destructive wipe that requires typing WIPE after biometric re-check.
Each MCP server in your registry has four axes. Base AiMANAC reads them but does not edit them. Satcheli is where they're tuned.
Where (data flow ceiling)
Classification ceiling. LOCAL / LAN / CLOUD. Refuses dispatch above the ceiling regardless of caller.
Default · CLOUDBy whom (caller policy)
Which callers may dispatch this server. HELM, OpenAI, Anthropic, Gemini, Local Ollama, etc. Server-level — per-tool granularity means "split into separate servers."
Default · All callers allowedHow loaded
Eager, lazy, or on-demand. Lazy lets a server appear in the registry but skips actual connection until first invoke.
Default · LazyHow called
Per-tool grants (allow/deny) plus per-server butler config (on/off, cadence). Butlers handle health poll, tools-cache refresh, auth pre-rotate.
Default · All tools allowed · Butler offExtra “where it lives” labels for compliance teams may arrive in a future AiMANAC update. Today the pasted URL already tells the system whether traffic should stay on your machine, your backend, or a third-party host.
One sign-in. Satcheli reads the same iCloud keychain AiMANAC writes to. You don't sign in twice. Biometrics gate the same session, and changes to your registry travel with that session.
Two App Groups, on purpose. One group holds your AiMANAC sign-in (the trust mesh). A second group holds the MCP registry data. Both apps see both groups. The split means a future sibling app could read the registry without ever seeing your sign-in, which is the right shape for things like a future read-only MCP browser.
Your backend stays canonical. When you add, edit, or remove a server in Satcheli, the change goes to your customer backend first; the App Group cache updates from there. So if you reinstall an app, the registry comes back from your backend, not from local state.
Approvals live in AiMANAC. Satcheli is where you set policy — what each server is allowed to do and which callers can use it. Live approval prompts (when an LLM tries to invoke a server) still surface in AiMANAC. Satcheli sets the rules; AiMANAC enforces them in the moment.
- Have AiMANAC installed first. Satcheli reads the trust-mesh keychain AiMANAC fills on first secure session. If you don't have AiMANAC yet, start with /apps/aimanac/.
- Buy Satcheli on the App Store. Separate listing, $1 one-time. Different App Store SKU, same Apple ID.
- Open AiMANAC → MCP cockpit → Open Satcheli. AiMANAC hands you straight into the server list. Satcheli reads the shared keychain and confirms pairing — no second sign-in.
- Absorb your first manifest. Paste a standard MCP manifest, import a file, or pull from Smithery / Anthropic Directory. Set the four axes per server. Toggle a butler if you want background polling.
No Satcheli server SKU. There is no satcheli-server container — do not run docker pull for this app. When you need a customer backend, deploy aimanac-server per /setup/llm/ (fleet co-host on one VM is fine).
A single Discord covers customer support and feedback across every app in the AiMANAC family — Satcheli included. We do not run separate forums, Typeform surveys, or in-app feedback chrome.
Support
Stuck pairing with AiMANAC, weird MCP server behaviour, or a butler that won't poll? Ask in Discord; the maintainer and other operators are in there.
Invite link pendingFeedback
Catalog source ideas, axis edge cases, suite-app integrations — all of it goes through Discord, openly, without ticket gating.
Invite link pendingDirect contact while the invite link is being
finalized: info@llmtech.net.