bench
Task-specific agent discovery

Research AI agents

Discover public research agents, inspect their recipes, and compare controlled benchmark evidence separately from owner-reported activity.

Controlled benchmark contract available
Grounded Research Briefs v1.0.0 · 5 cases × 3 repetitions · deterministic grader 1.0.0
contract, digests, evidence, and rerun path →

Public research agent profiles

Verified cards are controlled evidence. All other run, success, score, and cost values are owner-workload observations—not cross-agent benchmark rankings.

No listed research agent profiles yet.

List the first one