Mount Tally inside your AI.

Five steps, about two minutes. When you're done, the chat where you already research trades can pre-register theses, log evidence, and mint receipts — while Tally keeps the record neither you nor the model can rewrite.

01

Add the connector

Pick your AI — the steps adapt. In ChatGPT it's one click from the app directory; elsewhere the steps tell you exactly where the URL goes. Signing in happens in step 02.

1 · Copy this URL
https://tally.markets/api/mcp
2 · Open Claude and paste it

Add custom connector → name it Tally → paste the URL. Leave the OAuth fields empty — Tally registers itself.

Open Claude & paste
02

Sign in and approve

Your AI opens Tally's sign-in page. Signing in binds the connector to your journal, and only yours. Read permissions are safe to set to “always allow”; keep write permissions on approval if you want a human click before anything changes the record.
03

Say hello

In a new chat with Tally enabled:
Get started with Tally, then run my checkup.
Your AI reads the protocol and briefs you: active theses and their deadlines, plus your closed history and lifetime R.
04

Install the skill (optional, recommended)

The skill makes your AI proactive instead of polite: float a trade idea and it offers to pre-register a thesis; talk about “taking profits” and it checks your written rules first. Download the skill file, then: in Claude, upload it under Settings → Skills. In ChatGPT, add it to a Project's files — it loads in every chat inside that project.
05

Schedule the morning checkup

Add a daily scheduled task (Claude scheduled tasks, or ChatGPT Tasks) with exactly this prompt:
Run my Tally checkup, gather fresh readings for each active thesis's criteria, log them with record_reading, and flag any criterion that has objectively fired.
That one schedule runs the discipline loop without you: fresh evidence logged, deadlines surfaced, triggered rules flagged before you've had coffee.

Step 02 needs a Tally account — free, no card.

Tally never executes trades, and your AI only ever sees your own journal.

For security review

What an administrator evaluating this connector needs to know. Every claim here is checkable against the connector itself.

Endpoint
A single hosted endpoint, https://tally.markets/api/mcp, over streamable HTTP. Same URL for every user; identity comes from the access token, never the URL. There is nothing to install and no local process.
Authentication
OAuth 2.0 with dynamic client registration and PKCE — no shared secret to distribute. Consent is hosted at accounts.tally.markets. Requested scopes are openid, profile, and email, and nothing else.
What it can reach
Every tool is scoped to the authenticated user's own journal. The server reaches no third-party service on the user's behalf, browses nothing, and cannot read another account's data. It never queries the AI client's memory, chat history, or files.
It cannot trade
Read-only toward markets by architecture. No tool places, modifies, or cancels an order on any venue, and none moves money or crypto. Where a brokerage is linked, it is through SnapTrade with a read-only connection — broker credentials never reach Tally.
Tools
Ten, split five read and five write. Reads carry readOnlyHint and run without per-call confirmation; the three one-way operations — arming a thesis, firing a criterion, closing a thesis — are marked destructive so the client always asks first. Full inventory and annotations are readable from the connector.
Data and retention
What the user writes: theses, criteria, readings, positions, notes. Any venue session key is encrypted at rest (AES-256-GCM) before it touches the database. Receipts are public only for theses the owner has closed. Export or deletion on request at support@tally.markets. Full detail in the privacy policy.
Revoking access
The grant is per user and can be revoked at any time from the AI client's connector settings, which invalidates the token immediately. Revoking does not delete the journal.
Operator
Tally is operated by Prosperity Labs, LLC. Support at support@tally.markets, with terms and privacy policy published.