Treeship
CLI reference

match

Find agents by exercised, verified evidence — the Hub proposes candidates from its index; every record is re-verified on your machine before ranking.

treeship match finds agents that have actually run tools matching a glob, from their signed session.v1 records. The Hub's index proposes; your machine re-verifies each record against your pinned trust roots and re-checks the glob locally — a hub cannot inflate a candidate's rank with records that don't verify.

Usage

treeship match --hub <url> --exercised <glob> [--class <class>] [--min-sessions <n>]
OptionDescription
--hub <URL>Required. The Hub index to query
--exercised <GLOB>Required. Tool glob the agent must have exercised, e.g. "payments.*", "Bash(git:*)"
--class <CLASS>Restrict to self | runtime | countersigned evidence
--min-sessions <N>Minimum matching sessions (default 1)
--format jsonMachine-readable output

What "verified" means here

For each proposed candidate, every record must: signature-verify against your trust roots, be a session.v1 whose actor is the candidate, and exercise a tool matching your glob — recomputed locally, never trusting the hub's match. Candidates rank by locally verified session count; the output shows both numbers, so an index that over-proposes is visible:

agent://deployer     4/4 verified    payments.charge, payments.refund
agent://other-bot    0/3 verified    unverified (records did not verify against your trust roots)

JSON output

{
  "exercised": "payments.*",
  "count": 2,
  "candidates": [
    {
      "agent": "agent://deployer",
      "matched_tools": ["payments.charge", "payments.refund"],
      "proposed_sessions": 4,
      "verified_sessions": 4,
      "verified": true
    }
  ]
}

Informational command — exit code 0 either way. See GET /v1/agents/match for the Hub side, and Work history for the records themselves.