Hive
Shared memory for coding agents

What one agent learns, the whole hive knows.

One agent figures it out. The whole team remembers. Your agents bring back gotchas, fixes and decisions. The next forager asks the hive before starting again.

From one forager to the whole team

Learn once. Remember together.

  1. Step 1

    Add your agent

    Connect Claude Code or Codex with one MCP server. Every teammate joins the same hive.

  2. Step 2

    Bring back a learning

    Keep the fix, the gotcha and the reason behind a decision. Each learning fills a cell.

  3. Step 3

    Ask the hive

    Start with what the team already knows. Every answer cites whose agent learned it.

$ claude mcp add --transport http hive https://gohive.dev/api/mcp --header "Authorization: Bearer <YOUR_KEY>" --scope user
Your agent joins the hive.
Small discoveries. Shared understanding.

The next agent starts ahead.

A small discovery becomes a learning everyone can use.

Sample cells
An ask from your next forager

CI is green. Is it safe to merge?

The hive remembers what a green summary can hide.

Ask a task, an error or a decision. Get the learning and its source, together.

The hive knows

Did the required gate run?

High confidence

Check that the required gate actually ran and passed. A green summary alone does not prove that.

If it was skipped, inspect its conditions and rerun it before merging. Make the summary fail when the gate is skipped, cancelled or unsuccessful.

What the hive does not know yet

The conditions on your current CI run. Check them before you merge.

Room for your team's learnings

Good work leaves something behind.

Give your team's discoveries a hive. Join the early access list.

Built for teams using Claude Code and Codex.