Docs

Yoshi is one journal that any AI can write to. Connect a chatbot, give your agent the MCP server, or script it with the yoshi CLI.

Connect ChatGPT & Claude
Coding agents
MCP tools
CLI
Import & export

Connectors and the CLI need a Cloud account, because they reach your journal through the server. The apps work without one.

Connect ChatGPT & Claude

The MCP server is at:

https://api.getyoshi.app/mcp

Claude

Settings → Connectors → Add custom connector.
Paste the URL above.
Sign in to Yoshi and approve. The tools show up in your chats.

ChatGPT

Settings → Apps & Connectors → Advanced → turn on Developer mode.
Create a connector with the URL above and OAuth authentication.
Sign in to Yoshi and approve.

Then just say it:

"Lunch was a bowl of beef noodles. Save it to my Yoshi."
"Pull-ups 10 8 6 5, log it."
"Add 'ship the landing page' to tomorrow under Yoshi."
"How much protein did I eat this week?"
"What did we work on today? Make it my to-do list."

Coding agents

Any agent that speaks MCP over HTTP can use the same server. It signs in through your browser the first time.

Claude Code

claude mcp add --transport http yoshi https://api.getyoshi.app/mcp

Then run /mcp in Claude Code to sign in.

Codex

codex mcp add yoshi --url https://api.getyoshi.app/mcp
codex mcp login yoshi

Your own agent

The server speaks Streamable HTTP and uses OAuth 2.1 with dynamic client registration, so standard MCP clients sign in on their own. Discovery is at /.well-known/oauth-authorization-server.

MCP tools

Tools that write need the write permission you approve on the consent page. A write that fails halfway leaves nothing behind.

Tool	What it does
`log_food`	Saves a meal, with the AI's best guess of calories, protein, carbs, fat and fiber. Answers with the day's totals against your goals.
`edit_food`	Changes or removes one food item: its numbers, amount, name or meal.
`add_todo`	Adds a to-do to a day, optionally under a topic.
`check_todo`	Checks off (or unchecks) the to-do that best matches the text.
`log_training`	Saves a workout, a line an exercise, written the way you'd jot it: `Pull-ups 10 8 6 5`.
`edit_training`	Rewrites a day's training, to fix a number or remove an exercise.
`rename_exercise`	Renames an exercise across many days, so "pullups" and "pull ups" become "Pull-ups".
`get_day`	Reads a day: to-dos, training, food and totals.
`get_range`	Reads up to 92 days with averages, for "how often did I train this month".
`search`	Finds to-dos, food, training and notes containing some words, newest first.
`import_rows`	Imports a few hundred rows of old logs, in the CSV format's shapes.

CLI

yoshi is a single binary for signing in, importing and exporting.

yoshi login                              # opens the browser to sign in
yoshi whoami
yoshi import --dry-run todos.csv food.csv  # checks the files, changes nothing
yoshi import todos.csv food.csv
yoshi export --format md --out journal/  # a Markdown file a week
yoshi export --format csv --out backup/  # four CSV files
yoshi logout                             # signs this computer out

Sign-in: the token is kept in ~/.config/yoshi/config.json, readable only by you. Set YOSHI_TOKEN to use another, for scripts.
File kind: taken from the file's name (todos, food, training, notes), or set with --kind.
Exit codes: 0 worked, 1 failed or the import found problems (each printed as file:line: what), 2 wrong arguments.

Import & export

Yoshi takes a journal as up to four CSV files: todos, food, training and notes. Turn your old logs from Notion, Apple Notes, a spreadsheet or another app into these files, often by asking an AI to, then load them with the CLI or in the app (Settings → Data → Import CSV).

Only date (YYYY-MM-DD) is required. Unknown columns are ignored.
Importing twice doesn't duplicate: a row with an id replaces what was imported with that id before.
CSV export includes each row's id, so it imports back as itself.

The full format, every column of every file, is in import.md. It's written so you can hand it to an AI along with your old data.