Overview
Connect Claude or another agent to Walletguide over MCP at mcp.walletguide.com, read-only by default.
MCP is how an agent gets tools. Connect Walletguide and Claude can read your assets, search your transactions and ask for a real net-worth figure, using the same routes the app uses.
There is one server, and it is hosted:
https://mcp.walletguide.comThe earlier address, https://api.walletguide.com/mcp, keeps working for the connectors already using it.
Connect a client
One URL, browser sign-in, no local process and no key to store.
The tools
What an agent can call, how tools are named, and which are destructive.
Nothing to install and nothing to keep up to date: the tool catalogue is generated from the same OpenAPI document as the API, so a route added to the app reaches your agent on the next deploy. Sign in through your browser and the grant is revocable from your user settings. If your client cannot open a browser, it can send a wg_ API key as a bearer token to the same URL.
Read-only by default
Sign in through your browser and the consent screen offers write as a box, unticked for an existing account and ticked for one you just created to connect. Leave it unticked and the connection has only the read tools and no way to change anything.
That is not only caution. The read-only catalogue is less than half the size, roughly 17k tokens against 40k, and that is context the model spends before it has read a single number.
A grant that carries only walletguide:read is never shown a mutating tool. It is not offered one and refused on use; it does not appear.
An API key carries its own permissions instead of scopes, and is shown the tools those permissions allow. A viewer key, which is what a new key is by default, gets the read-only catalogue and an editor key the data writes as well. Wallet settings and bank connections appear only for a key whose permission list names wallet:manage or integrations:manage, and no key is shown wallets_create or the access requests. The counts per connection are in the tools page.
Setting a wallet up
An agent can do more than read. With the person's consent it can create their wallet, hand them a link to connect their banks, and do the setup work; changes to members, API keys and deleting a wallet it can only ask for, and the person approves each one in the app. Connect a client has the details, and an agent asked to set Walletguide up can follow the step-by-step guide at walletguide.com/agents.md. People connecting their own assistant start at walletguide.com/agents#connect.
Tools
One per route, named after it. GET /wallets/{walletId}/assets is assets_list, POST /wallets/{walletId}/transactions/search is transactions_search, and the metric slices are data_netWorthTimeSeries and friends. The endpoint reference describes all of them; a tool takes the same parameters as its route.
Anything that deletes or overwrites data carries a destructive annotation, updates included, so a client that asks before running one will ask.
Getting good answers
Ask a data_* tool for a metric. It computes over the full history on the server and applies the rules raw rows do not carry: splits, transfers between your own accounts, offset categories. An agent that lists transactions and adds them up gets one page of them and no warning. See metrics.
Amounts are decimal strings, and negative means money out. Ids are uuids, so find one with transactions_search or a list tool rather than guessing.