Hubo
A local, code-review agent framework where two distinct AI agents implement and review each other's work until every review finding is reconciled. Useful for demanding codebases that require adversarial review cycles. Each agent generates, runs, edits, and re-evaluates until convergence, helping catch subtle bugs early and documenting rationale. Open-source under the MIT license; batteries-included tooling for integration and logging.
Reader rating
No ratings yet
You might also like
Related tools
Zero by VM0 is your trustworthy AI teammate that connects to 100+ tools and completes real work hands‑free — from triage queues to Slack‑first reporting, CRM updates, and research tasks. Focuses on consistent outputs, low latency runs, and defensible agent policies so you can deploy it to production without risking burn‑out or missteps. Offered as cloud‑hosted or self‑hosted.
A security scanner and methodology for hardening AI agent frameworks against ‘consequential’ actions that can fire without approval checks. Scans file trees and configuration for a curated list of 25 risky actions (write/commit/db calls/file uploads/email), ranks false positives, and outputs a remediation strategy so teams can safely roll out agents. Delivered as self-hosted CLI and server — lightweight and policy-driven.
A governed graph layer that turns disparate data sources (databases, APIs, codebases) into a unified, queryable graph for AI agents. Acts as a read-only MCP gateway that enforces security, provenance, and topology constraints so agents only traverse edges and nodes allowed by policy. Coffee‑script codebase; backed by GB-sized reference tests and open to integrations.