Stop hopping between
every client's workspace.
ProxyLoom bundles each client's Slack, Notion, Google and Miro behind a single connection for Claude, ChatGPT or any MCP-ready AI agent. “Pull last week's decisions from client A's Slack, then add today's notes to client B's Notion” — one instruction spanning engagements, with no re-authentication.
No credit card. Free while in beta.
Slack3 connections
Miro2 connections
Google Drive2 connections
Notion1 connections
- Northgate TradingLive
- Ridgeline ManufacturingLive
- Fairview CreativeLive
- Northgate TradingLive
- Ridgeline ManufacturingLive
- Fairview CreativeLive
- Northgate TradingLive
- Ridgeline ManufacturingLive
Five engagements means five workspaces
The bottleneck is not the thinking — it is the switching between client screens around it.
One workspace per client
Client-issued Slack, Notion and Google accounts, one set per engagement, switched by hand all day. The more guest workspaces you sit in, the more it costs.
AI agents hold one account per service
Claude, ChatGPT — every agent connects to one Slack at a time. Connecting client B evicts client A, so a question that spans engagements simply cannot be asked.
Client information must not mix
Client A's material cannot leak into client B's work. Splitting the AI environment per engagement fixes that — and makes you the one who has to remember everything.
Connect each client once, then ask by name
ProxyLoom holds every client connection and presents them to your AI agent as one. The agent picks which client's connection to use from the wording of your request, so client B's connection is never used for client A's work.
- 01
Authorize each client's Slack, Notion, Google or Miro
Connect from the dashboard, one client at a time. Each connection gets a routing key derived from the workspace you authorized, and you can name it yourself — either is how you refer to it.
- 02
Register one connector in Claude
One ProxyLoom URL, one click from the dashboard. Every client you connected is reachable through it.
- 03
Ask by client name
“In client A's Slack…” — the model picks the right client's connection from context.
How the week actually runs
Say the client's name and the AI picks the right connection. The asks look like this.
“Pull last week's decisions from client A's Slack project channel.”
Catching up on an engagement without scrolling the channel yourself — and the next client is one name away.
“Append today's meeting notes for client A to their minutes page in Notion.”
Every client keeps minutes somewhere different. Remembering where stops being your job; you just say the name.
“Across client B's and client C's Notion, pick up the pages updated this week and give me one digest.”
Cross-engagement status in one ask — possible only because there is one connection. Weekly reporting prep ends here.
“Read the sticky notes on client A's Miro board from last week's workshop and summarize them by theme.”
Post-workshop synthesis straight from the board. Different Miro accounts per client, one name to point at.
It holds your tokens, so it is built like it
A proxy that stores credentials for every account you own has to earn that.
Credentials encrypted at rest, per customer
Stored tokens are encrypted with keys that differ per customer, and decrypted in memory only at the moment of an upstream call.
Per-customer isolation enforced by the database
Separation between customers is enforced at the database layer, not by application code — so even a bug in our code cannot reach another customer's data.
We never look at relayed traffic
ProxyLoom never reads the content of the traffic it relays between your AI agent and your SaaS accounts — never inspected, never stored, never logged. The audit log records metadata only: which connection, which tool, when.
Deleting a connection revokes the stored token
Removing a connection deletes the stored token from ProxyLoom and sends a revocation request to the SaaS, so the credential is no longer valid on either side. *Limited to services that provide a revocation API.
Hosted in Japan
ProxyLoom runs entirely on infrastructure located in Japan, and credentials and audit records never leave the country.
Free while in beta
ProxyLoom is in private beta. Every feature is free while the beta lasts.
Works with the AI you already use
…and any other MCP-compatible client
Product names and logos are trademarks of their respective owners.
Frequently asked questions
Which MCP clients work with it?
Claude, Codex, and any other client that speaks Streamable HTTP MCP and OAuth 2.1. Registration is a single ProxyLoom URL — the proxy handles discovery and consent from there.
How does the model know which account to use?
Each connection gets a routing key derived from the workspace or business you authorized, and you can give it a name of your own in the dashboard (e.g. “Northgate Trading”). Every relayed tool gains an account argument, and the model picks the connection from context using either name.
Is my data stored?
Not the payload. ProxyLoom never inspects the messages and tool calls it relays, and never writes them to logs. What it does store is your encrypted tokens and audit-log metadata — connection, tool, timestamp. Tokens are decrypted in memory just for the duration of an upstream call.
What happens if I disconnect an account?
The stored token is deleted from ProxyLoom, and revoked at the SaaS where the service provides a revocation API (Slack, Google, …). No credential of yours stays behind in ProxyLoom.