|
ROJO · Behavioral security for AI agents
Rojo is behavioral security for AI agents. We drive your running agent with adversarial inputs, record every tool call it actually makes, and hand you the ex...
|
84 % |
|
Primary Path: approvals and a provable audit trail for AI agents
Self-hosted approval inbox and tamper-evident audit trail for AI agents that take real actions. Open source, Apache-2.0.
|
83 % |
|
AI Agent Red Teaming & Prompt Injection Testing — okjul
Adversarial testing for AI agents. okjul red-teams your agent for prompt injection and tool misuse, then hands you reproducible evidence auditors accept.
|
82 % |
|
positif · the agent control plane
Know what your AI agents are doing. Prove it later. Positif permission-checks, logs, and loop-protects every request between your agents, without ever readin...
|
81 % |
|
Keel — AI agents you can undo
|
81 % |
|
Arcezia: check AI agent tool calls before they run
Arcezia checks an AI agent's tool call, and the facts behind it, before it runs, then answers ALLOW, BLOCK or REVIEW with a signed record.
|
81 % |
|
Veractly — Let agents act. Know the verified outcome.
Verify consequential agent actions across business systems. Apply policy, inspect the evidence, and recover when the external outcome diverges across your st...
|
81 % |
|
Mnestiq: forensic evidence for AI agents
Trace any network connection back to the AI agent step that made it, and prove the record wasn't altered. Open-source, tamper-evident evidence for AI agents.
|
81 % |
|
AI Agent Regression Testing for Tool-Using Agents | Konsista
Catch AI agent regressions before release. Repeat mocked tool-use scenarios to expose skipped approvals, wrong targets, missing tools, and duplicate actions.
|
80 % |
|
ABERNATE · permission to ship the agents
Finished agents sit in staging because nobody will sign for them. ABERNATE decides whether each action is permitted at the moment it happens, and writes a si...
|
80 % |
|
UluOps — The operations layer for agentic work
UluOps turns agent runs into tracked findings — scored, fingerprinted, and carried forward until they're resolved. Git remembers what changed. UluOps remembe...
|
80 % |
|
AI Agent Gatekeeper: Your Agents Act, You Stay in Control | Clawvisor
Your AI agent needs access to your email, calendar, and code. Clawvisor makes sure it stays on track. Purpose verification, credential vaulting, and audit lo...
|
80 % |
|
dogud — AI agent orchestration that finishes the job
dogud turns plain instructions into coordinated agent workflows — planned, executed, and verified without babysitting.
|
80 % |
|
Agent Centipede — a local-first agent harness that finishes things
Agent Centipede is a local-first agent harness. Tell it what you need done; it carries the follow-through and hands you a receipt that shows it actually fini...
|
80 % |
|
Sentrik — Authorization Infrastructure for AI Agents
Sentrik decides what an AI agent is allowed to do — before it acts. Every consequential action is checked against your policy and authority, then recorded.
|
80 % |
|
AI agent governance, enforced on the device | kriya
AI agent governance on the device: deny-by-default policy before every action, human approval for risky ones, signed receipts your auditor re-verifies offline.
|
80 % |
|
Tafadzwa Mukungurutse · agents fail silently; the gate catches it
I build the checks between a working demo and an agent you can leave running: deterministic assertions, bounded retries, human approval. And I write down wha...
|
80 % |
|
VERAX · The black box for your AI agents
VERAX turns every decision an AI agent makes into a signed record: what was asked, which rule applied, what happened. The record stays on your machine. Conar...
|
80 % |
|
Stonefold — control for AI agents that act on real systems
Companies are starting to let AI agents take real actions in their systems. Stonefold is the checkpoint that keeps those actions inside rules your organizati...
|
80 % |
|
Coban — a desktop workbench for coding agents (Codex, Claude Code, Antigravity)
Coban runs Codex, Claude Code and Antigravity on your Mac as one workbench: threads, agent teams with a Lead, review bound to the exact diff and tests, appro...
|
80 % |
Detected social media networks