Tools Reference

FreeRide gives your AI agent a set of tools, organized by what they're for. You don't call these directly — your agent uses them based on your conversation. This reference helps you understand what's available and when each tool is used.

Any app, including chat

Tools that work everywhere, with no repo and no session.

Tool	What it does
catch_up	A short brief on a project: what it is, recent work, what's planned, recent decisions, how many open ideas. In chat apps that support it, it shows as a card. "Catch me up on Mealwise"
list_projects	Your projects, most recently worked on first. Used when it isn't clear which project you mean; in chat it shows a picker.
save	Saves new ideas and decisions to a project in one call, and shows a receipt with Undo in apps that support cards. "Save this idea to Mealwise"

The project is passed by name or id. If FreeRide can't tell which project you mean, it says so and the agent asks you. It never saves to a guessed project.

Session

Tools that manage a working conversation in a coding agent.

Tool	What it does
start_session	Opens a session and returns your project's current state — features, recent progress, open ideas, documentation index. Called automatically at the start of every conversation in a linked repo. Apps with small tool-output limits, like Codex, get a shorter version with the same facts.
end_session	Closes the session with a rich narrative summary on top of the structured completion log. Triggered by saying "end session." If you don't, FreeRide auto-closes idle sessions with a less-rich rollup.

Working on Features

Tools for the core work cycle: get context, do the work, log what happened.

Tool	What it does
begin_work	Loads context for a specific feature before implementation — open ideas, prior decisions, existing docs, file manifest, the features it's most connected to, and records on other features that mention it, so the agent checks them when a change reaches across. Use for substantial work.
complete_work	Logs meaningful progress and tags it to one or more features. The primary record of work done. Returns doc snapshots so the agent can check if documentation needs updating.
log_iteration	Adds a follow-up to work that was just logged, for example when you ask for changes. The refinement joins the same entry, so the history stays one story per piece of work.
confirm_docs	Records the agent's documentation decision after reviewing the doc snapshots from complete_work. Part of the doc freshness flow.
complete_feature	Marks a feature as done for the first time. Verifies an implementation doc exists, transitions the feature status, and logs a completion.

The typical cycle: begin_work → do the work → complete_work → confirm_docs. For small changes, complete_work alone is enough — the full cycle is for substantial implementation.

Creating & Capturing

Tools for adding new items to your project knowledge.

Tool	What it does
create	Creates a new feature, idea or project. Features give your project structure. Ideas capture thoughts — attached to a feature or project-level. A new feature can also be created as a part of an existing one. "Let's track this as a feature" or "Add this to the backlog" or "Start a new project for this"
capture	Records a decision or learning. Decisions capture a choice and the reasoning behind it — architecture, scope, direction, anything worth remembering (alternatives optional). Learnings preserve insights worth keeping. "Let's go with Postgres because..." or "TIL WebSockets don't work behind..."

Knowledge & Documents

Tools for maintaining your project's documentation.

Tool	What it does
save_document	Creates or updates a knowledge document. Seven types: overview, implementation, architecture, guide, backlog, UX flow, vision. Multiple docs per feature supported.
patch_document	Applies targeted edits to an existing document — small changes without resending the full content.
search	Queries across your entire project — features, documents, ideas, completions, decisions, learnings, files. Full-text search with filters by feature, date, status, priority, and more. Very long documents come back in parts.

Updating & Logging

Tool	What it does
update	Updates features or ideas — status, priority, description, title. Also handles feature status transitions (backlog → in progress → done), and moves a feature into another feature as a part, or back out on its own.
log_progress	Logs a mid-work update, blocker, or insight without marking a full completion.

Helpers

Two tools you'll rarely notice: continue_chunked fetches the next part of a long result, and undo_save is what a save receipt's Undo button calls (it works for a couple of hours after saving).

How You Interact With These

Most of the time, you don't. The agent calls these tools on its own — opening a session at the start, logging completions after meaningful work, capturing decisions as they come up in conversation. That's the point: structured project knowledge as a byproduct of the work, not a chore on top of it.

When you do want to nudge it, plain language is enough — the agent maps loose phrasing to the right tool:

• "Catch me up on this project" → catch_up
• "Track Authentication as a feature" → create
• "That's a decision worth keeping" → capture (or save in chat)
• "What did we figure out about the database last week?" → search
• "The implementation doc needs a refresh" → save_document or patch_document
• "We knocked that one off already" → update
• "Put Parser under Dunkman" or "Take Kaizershop out of Clients" → update
• "Wrap up what we just built" → complete_work
• "What should I work on next?" → start_session context + search

You don't have to think in tool names.

Related

• How It Works → — the session flow these tools sit inside
• Apps & Agents → — which apps use which tools
• Dashboard Guide → — where the data these tools capture surfaces