Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Toki Toki no Mi Record: rejected Redis due to lack of persistence."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An MCP server that remembers why each decision was made — a local, searchable decision log.
A Model Context Protocol (MCP) server — one focused tool you plug into an MCP-compatible client (Claude Desktop, Claude Code, Cursor, and others).
A conversation's context window forgets. A few sessions later, the reasoning behind an architectural choice is gone, and you re-litigate decisions you already settled — or quietly reverse them without noticing.
Toki Toki no Mi is a causality ledger. It records each decision, the alternatives you rejected, and why you chose as you did, then lets your assistant retrieve that history later — and flag a new proposal that contradicts a past decision.
Related MCP server: Decision State MCP
Two tools:
log_causality_point — record a decision, the rejected alternatives, and the justification
time_leap_retrieval — recall past decisions (all of them, or filtered by keyword)
Decisions are stored locally in SQLite under ~/.grandline/ and also mirrored to
a Markdown file, so any external search tool can be pointed at the folder to read
them.
Tool reference: log_causality_point(decision_made, alternatives_rejected, justification) · time_leap_retrieval(query: str = "")
Self-contained, no API key, fully local. The ledger is a single SQLite file on your machine with no external dependencies. Nothing is sent anywhere.
Requires Python 3.10 or newer.
pip install toki-toki-mcpThis installs the toki-toki command.
Then register it with your MCP client — either run claude mcp add toki-toki -- toki-toki,
or add this to the client's config (e.g. claude_desktop_config.json):
{
"mcpServers": {
"toki-toki": { "command": "toki-toki" }
}
}Restart the client and the tool is available.
Toki Toki no Mi is part of a four-tool set. Each tool stands alone, but installing all of them unlocks a final surprise via Laugh Tale.
Package | ||
🗿 | Road Poneglyph | |
⚔️ | Conqueror's Haki | |
⏳ | Toki Toki no Mi · ← this repo | |
🏴☠️ | Laugh Tale |
git clone https://github.com/xfloukiex-lab/toki-toki
cd toki-toki
python -m venv .venv && . .venv/Scripts/activate # macOS/Linux: source .venv/bin/activate
pip install -e ".[dev]"
python tests/test_toki_toki.pyReleased under the Apache-2.0 license. The icon is original artwork
(see icon.svg), generated locally; the repository ships no third-party assets.
This is a fan-inspired project. It is not affiliated with, sponsored by, or endorsed by the creators or rights holders of One Piece (Eiichiro Oda / Shueisha / Toei Animation). "One Piece" and related names are used only as thematic flavour and remain the property of their respective owners.
Anchor a moment in time. Record a project decision, the alternatives rejected, and why it was chosen — for future historical reference.
Args: decision_made: The final path / feature / architecture settled on. alternatives_rejected: What other options were considered and discarded. justification: Why this decision was made over the alternatives.
Inputs: decision_made (required), alternatives_rejected (required), justification (required). Returns: result
Leap backward into the timeline. Retrieve past decisions and the reasons behind them to understand how the current state came to be.
Args: query: Optional keyword to find a specific past decision. Empty returns the 10 most recent anchors.
Inputs: query. Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 2 tools
The two tools have clearly distinct purposes: one records decisions, the other retrieves them. No overlap or ambiguity.
Both tools use a consistent verb_noun pattern with snake_case (log_causality_point, time_leap_retrieval), following a predictable convention.
With only 2 tools, the server is minimal but focused. While it covers the core operations of logging and retrieving decisions, more tools could be expected for a broader time-travel theme.
The set covers create (log) and read (retrieval) but lacks update and delete operations for decisions. This is a notable gap for full lifecycle management.
Markdown-based note-taking with a hosted MCP server. Your notes serve you and your AI.
MCP server for generating rough-draft project plans from natural-language prompts.
Hosted MCP memory: save sessions/decisions once, search from Claude, Cursor, ChatGPT. EU-hosted FTS.
One memory, every AI. A shared, user-owned markdown memory your AI clients read and write over MCP.
