Integrations

Using Karma Yogi with Claude

Karma Yogi ships a remote MCP (Model Context Protocol) connector. Link your account to Claude and ask about your own study data in plain language — no app switching.

Server URL

https://karma-yogi-api.onrender.com/mcp

1What it is

The connector links your Karma Yogi account to Claude on claude.ai. Once connected you can ask things like “analyse my last 20 mocks”, “what's my weekly goal progress”, or “log a 45-minute Quant session” right inside a claude.ai conversation.

2How to connect

In claude.ai, open Settings → Connectors and choose to add a custom connector.
Enter the server URL above.
claude.ai registers itself automatically (Dynamic Client Registration — no client ID or secret to type in) and redirects you to a Karma Yogi sign-in/connect screen.
Sign in with your existing Karma Yogi account. If you're already signed in, you'll see which app wants to connect and what it can access — review and approve.
You're redirected back to claude.ai, connected.

Important: this flow only works when started from Claude. Visiting the connect link directly shows a dead-end message telling you to start from Claude — there's nothing to connect to.

3What Claude can do

Reading your data

list_sessions

Lists your logged study sessions (study, analysis, or test), optionally filtered by date range, most recent first. Includes each session's revision round if it has one.

e.g. “What did I study this week?”

list_subjects

Lists all the subjects you've created, with their name, color, and icon.

e.g. “What subjects do I have set up?”

list_full_mocks

Lists your logged full-length mock tests — provider, section-wise and overall scores/percentiles, tags, and notes, most recent first.

e.g. “Analyse my last 20 mocks.”

list_sectional_tests

Lists your logged sectional tests — provider, section, score/percentile, tags, and notes, most recent first. Section names come from your own subjects and vary by exam.

e.g. “How have my Physics sectionals trended?”

get_goals

Gets your target exam (name and date, if you've set one) and your weekly study-time goals.

e.g. “How many days until my exam?”

get_study_stats

Gets aggregate stats — total time studied, session count, average and longest session, average mood, active days, and this calendar week's minutes.

e.g. “How am I doing against my weekly goal?”

Logging new data

create_subject

Creates a new subject. If one with that name already exists, it reuses it instead of duplicating.

e.g. “Add a new subject called Reading Comprehension.”

log_study_session

Logs a completed study session (subject, duration, optional topic/mood/timestamp). Auto-creates the subject if it doesn't exist yet, and won't create a duplicate if the same call is repeated.

e.g. “I just studied Quant for 45 minutes, log it.”

log_full_mock

Logs a full-length mock test result. Per-section results go in `sections`; section names come from your own subjects. `provider` and `durationMin` are both required — duration triggers the linked study session and revision (R0) stamp the app itself always creates, so a mock logged here matches one logged by hand. Won't duplicate a mock with the same name and date.

e.g. “Log today's IMS mock: Physics 60, Chemistry 55, Maths 70, overall 185, took 3 hours.”

log_sectional_test

Logs a sectional test result for one section of your exam. `provider` and `durationMin` are both required, same reasons as above. Won't duplicate a sectional with the same name, section, and date.

e.g. “Log a 40-minute IMS Chemistry sectional I took today, scored 28.”

Correcting data you already logged

Claude looks the record up first, then sends only the fields that change — anything it doesn't mention is left exactly as it was, so fixing one score won't wipe your notes.

update_study_session

Corrects a logged session's topic, duration, mood, start time, subject or revision round. Only the fields Claude passes change. Won't move a session to a subject that doesn't exist, and can't re-link it to a different test.

e.g. “That Quant session yesterday was 90 minutes, not 45.”

update_full_mock

Corrects a full mock. Per-section edits only touch the sections named — fixing Quant leaves VARC, DILR and Quant's own notes alone. A section the mock doesn't already have is rejected rather than added.

e.g. “In SIMCAT 7, my Quant score was 35 not 30.”

update_sectional_test

Corrects a sectional's scores, percentile, duration, tags or notes. Everything not mentioned is preserved.

e.g. “Fix that QA sectional — percentile was 96.”

Quick-start prompts

Karma Yogi registers three canned prompts that clients like claude.ai surface as one-click starters, so you don't have to type them out:

Analyze mocks

Look at your recent full mocks and sectionals, identify weak sections/topics and trends, and suggest 2–3 focus areas

Weekly progress

Check your weekly study-time goal progress and exam countdown

Revision due

Work out which mocks/sectionals are due for a spaced-repetition revision round, and how overdue

4What Claude cannot do

Cannot delete anything — nothing you've logged can be removed through Claude.
Cannot edit subjects, goals, or preferences — those shape how everything else is interpreted, so they stay app-only. Sessions, mocks and sectionals can be edited.
Cannot create a subject as a side effect of an edit — asking Claude to move a session to a subject that doesn't exist is refused rather than inventing one.
Cannot access AI Insights, the AI expert chat, or consolidated mock notes.
Cannot see your friends' data, leaderboard entries, or any other user's data.
Cannot be used outside the 13 tools above — the access token is cryptographically scoped to the MCP connector and cannot authenticate against the regular app API.

5Example prompts

Analyse my last 20 mocks — what are my weak areas?
Which of my mocks and sectionals are due for a revision round, and when?
I just finished a 30-minute Physics session, log it.
How am I tracking against my weekly study goal this week?
Add a new subject called 'Reading Comprehension'.

Revision timing is worked out by cross-referencing your session revision rounds against mock/sectional dates using the spaced-repetition schedule.

6Disconnecting

Go to Profile → Preferences → Connected Apps. Each connected app has a Disconnect button and a confirmation step.

Disconnecting revokes refresh tokens for that app. A still-valid access token keeps working until it expires — access tokens are short-lived (15 minutes by default), so access fully stops within about 15 minutes, not instantly. You can reconnect any time.

7Troubleshooting

First request after a while looks like it failed or timed out.: The API can go idle and cold-start — the first call after inactivity can take 10–30 seconds. Not an error; just try again.
Claude says the connection failed, or shows a 401-type error.: Your connection has likely expired or was disconnected. Reconnect using the steps above.
Claude says you're sending requests too fast.: Each tool call is capped at 30/minute and 400/hour per account, mainly to catch runaway loops. Wait the suggested number of seconds and retry.
Reconnecting after disconnecting.: Just repeat the connect steps — there's nothing to clean up first.

8Privacy

The connector only ever reads or writes your own Karma Yogi data — no access to any other user's information. Karma Yogi does not see what you type into your conversations with Claude; Anthropic's handling of that is governed by their privacy policy.