HOW IT WORKS
A build log in three steps
Keep local sessions local. Publish only the reviewed summary.
01
Connect an agent workflow
Claude Code, Codex, Cursor, or Antigravity. One CLI command starts the flow.
$ npx agentfeed init
02
Review the public summary
AgentFeed shows summary, metrics, and prompt fields after automatic redaction.
$ agentfeed preview --remote
03
Publish a worklog
Ship one reviewed card to the feed and manage private or unlisted visibility.
$ agentfeed publish --open-review
Build in public, with AI
Show what you shipped, not just how many tokens you spent.
Discover real agent workflows
See what other builders prompted, changed, tested, and shipped.
AI development portfolio
Each session becomes a searchable portfolio entry.
Answer engine ready
Fast answers about AI agent work sharing
Concise public answers help search and answer engines understand AgentFeed, privacy boundaries, supported agents, and portfolio use cases.
What is AgentFeed?
AgentFeed is a public worklog platform for AI-assisted development. It converts local agent sessions into reviewed cards that show outcomes, evidence, and privacy-safe metrics instead of raw transcripts or raw code.
Does AgentFeed upload raw code or private prompts?
No. The default flow uploads a reviewable worklog draft, redacts sensitive public fields, and keeps raw transcripts and raw diffs private unless a user explicitly changes privacy settings.
Which AI coding agents can AgentFeed collect from?
AgentFeed focuses on Claude Code, Codex, Cursor, and Antigravity session evidence, while keeping generic metadata paths for future agent tools.
How does AgentFeed help AI portfolios and answer engines?
Each public worklog creates a structured, searchable explanation of what changed, which agent helped, what evidence was collected, and what outcome was shipped.
Supported agents
Claude Code
Codex
Cursor
Antigravity CLI
+ more coming soon