Flatbread

io.github.flatbread-cliPublished by GitHub account flatbread-cliv0.9.8Updated Nov 1, 2026

Share a coding agent's session as one link, secrets redacted first; pull another run's brief.

VerifiedSTDIODesktop onlyKnowledge & MemoryDeveloper ToolsCommunication & Collaboration

Overview

AI-generated overview

Flatbread records AI coding-agent sessions as shareable links with secrets redacted, and lets an assistant search other runs and pull their briefs.

What you can do
  • Search recorded coding-agent runs, including your own and public runs, for one that already did a similar task
  • Fetch a run's execution brief: the request, human steering, blockers and fixes, test results and diff
  • Share the current session as a Flatbread run, with a review page that redacts secrets by default
  • Choose who can open a shared run: private, unlisted or public
When to use it
Useful when you want an assistant to check whether someone already solved a task or hit the same error before starting, to reuse another run's diff and guardrails, or to publish a finished session as one reviewable link.
Requirements
Runs locally over stdio as the npm package flatbread-cli, requiring Node.js 20 or newer. A one-time flatbread setup connects supported agents, and a Flatbread account sign-in is needed to confirm and share a run. The package name is flatbread-cli; the npm name flatbread belongs to an unrelated project.
Before you install
Sharing uploads session content to Flatbread once you confirm on the local review page; secrets are redacted by default, but you should still review the page and hide more words before confirming. Choosing Unlisted or Public makes the run readable by anyone with the link. Search covers public runs on flatbread.io as well as your own account's runs.

Installation

In SourceWeft

  1. Open Flatbread in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

Flatbread

Flatbread turns an agent session into one link: what was asked, every step, what broke, what fixed it and the diff. People read it. Agents pull it and pick up where it ended.

The package is flatbread-cli. Never run npx flatbread or npm install flatbread: that npm name belongs to an unrelated project.

Share your first run

  1. Install it: npm install -g flatbread-cli. You need Node.js 20 or newer (nodejs.org).
  2. Connect your agents, once: flatbread setup. Then you can share from their chat. Updates keep it current, and an agent you install later is connected the next time Flatbread runs, so you never run it again.
  3. When a session is done, say share this run in its chat, or run flatbread share in a terminal. From a terminal it says which session it took. Not the session you meant? It prints the command for every other tool's newest, such as flatbread share gemini.
  4. A review page opens in your browser. It runs on your own computer (127.0.0.1), and nothing has left it yet. No tab? Open the link the terminal or your agent printed.
  5. Check it. Keys, passwords and personal details are already hidden. Hide more words if you like, and choose who can open the run: Private (only you), Unlisted (anyone with the link) or Public.
  6. Press Confirm & share. Not signed in yet? Press Sign in to Flatbread first: it opens a new tab, and this page keeps what you chose.
  7. Copy your link: the page that follows shows it, with Copy link.

Rather have your agent do steps 1 and 2? Tell it: Install and set up Flatbread using https://flatbread.io/install.md

Anyone's agent can then say "pull flatbread " to read what that run learned.

Flatbread reads sessions from Claude Code, Codex, Cursor, OpenCode, Google Antigravity, Gemini CLI, Cline and GitHub Copilot CLI. flatbread setup connects Claude Code, Codex, Cursor, Google Antigravity, OpenCode, Gemini CLI, GitHub Copilot CLI and Cline.

In this repository

The docs, the changelog and the issue forms. The CLI is published on npm as flatbread-cli; its source is not published here.

Issues are for bugs, missed redactions and agents you want Flatbread to read. Site: https://flatbread.io

Source: README.md at commit 3637ca6

Tools

3

flatbread_search_traces

Search Flatbread runs

Search recorded AI coding-agent runs (your own Flatbread account's, and public runs on flatbread.io) for one that already did a similar task. Use it BEFORE starting a task that someone may have solved before, or after hitting an error, to find a run that hit the same wall. Returns run IDs; pass one to flatbread_get_brief for what that run learned.

Parametersqueryharness

flatbread_get_brief

Get a run's execution brief

Fetch the execution brief of one Flatbread run: what was asked, how the human steered it, each blocker and the fix (so you do not repeat it), what the tests said, and the diff. Every quote is the run's own text. Pass a run ID from flatbread_search_traces, or a flatbread.io/r/<id> link.

Parametersslugrequired

flatbread_share

Share this run

Share the session you are in now as a Flatbread run. Call it when the user says "share this run", "flatbread share", "share to flatbread" or anything like it. It finds this session, leaves the share request out, opens a review page with every secret redacted by default, and returns its link at once - give the user that link: it is uploaded when the user presses Confirm there, which shares it to their Flatbread account. That press is the user's to make, never yours. Always pass cwd and suggested_title. For only part of a long session ("share from prompt 4 to 8"), pass from and to_prompt.

Parameterscwdfromharnessto_promptredact_termssuggested_title

Version history

1
  1. v0.9.8LatestOct 31, 2026

Score

80% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts
  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listedRequired15 pts
  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

    Declare a license in the repository or the registry entry.