I built Flare because I kept losing track of what my agents were doing.
You hand Claude Code or Codex a task, come back twenty minutes later, and there are thirty changed files and a chat transcript. The transcript tells you what the agent said. It doesn't tell you that one of those files is imported by nine others, or that the one change nothing tests is the one holding the app together.
So Flare reads the repo instead of the conversation. Files become nodes, imports become edges, and the map updates as the agent writes. Every change is attributed so when two agents cross over the same file you see it as a crossing on the map, not as a merge conflict tomorrow.
A few things I'm glad I got right:
• The terminal is a real terminal. Bring your own agent. Flare doesn't wrap it, proxy it, or touch your keys. • Every change burst is snapshotted locally, so you can diff and revert one file or the whole tree without involving git. You also get notified by every risky change and have access to a "diff sub-graph" showing what changed in the last session. • The task board is exposed over MCP, so the agent picks up work and asks its questions there. You're both looking at the project rather than at each other's messages. • It all runs on your machine. No account, no telemetry, no cloud. MIT licensed.
I'd love to hear how you're keeping track of what your agents change right now, and if you try it, what the graph tells you about your own repo that you didn't already know.
PH 用户
Can multiple agents work on the same project at once while Flare shows their changes separately?
I built Flare because I kept losing track of what my agents were doing.
You hand Claude Code or Codex a task, come back twenty minutes later, and
there are thirty changed files and a chat transcript. The transcript tells
you what the agent said. It doesn't tell you that one of those files is
imported by nine others, or that the one change nothing tests is the one
holding the app together.
So Flare reads the repo instead of the conversation. Files become nodes,
imports become edges, and the map updates as the agent writes. Every change
is attributed so when two agents cross over the same
file you see it as a crossing on the map, not as a merge conflict tomorrow.
A few things I'm glad I got right:
• The terminal is a real terminal. Bring your own agent. Flare doesn't wrap
it, proxy it, or touch your keys.
• Every change burst is snapshotted locally, so you can diff and revert one
file or the whole tree without involving git. You also get notified by every risky change and have access to a "diff sub-graph" showing what changed in the last session.
• The task board is exposed over MCP, so the agent picks up work and asks its
questions there. You're both looking at the project rather than at each
other's messages.
• It all runs on your machine. No account, no telemetry, no cloud. MIT licensed.
I'd love to hear how you're keeping track of what your agents change right
now, and if you try it, what the graph tells you about your own repo that
you didn't already know.