|
On the Bench · AI tool benchmarks
On the Bench: reproducible AI tool benchmarks. Current category: self-hostable AI gateways, measured on the same box, same mock, same load, same CPU pin.
|
79 % |
|
VAIL — Scan your agents for vulnerabilities
VAIL scans your agents for behavioral vulnerabilities. It reads an agent's definitions, finds where its rules are too brittle or hard to follow, and probes t...
|
77 % |
|
Modelometer, the record of how AI models behave, and when they change
196,750 model calls recorded across 36 frontier AI endpoints from 7 providers. Behavioral baselines for 24 seats: mean stance on the 1–5 axis, refusal rates,...
|
77 % |
|
Learn how AI benchmarks work
|
77 % |
|
OpenResearch — The benchmark is the oracle
A decentralized, agent-driven research protocol. Code improvement is proof of work — measured by deterministic benchmarks, attested in hardware, rewarded on-...
|
77 % |
|
Lanier Developments — AI Engineering Consultancy
Lanier Developments builds production agentic systems and the platform discipline to run them. Every architecture decision is reviewed across two independent...
|
76 % |
|
The Resonance Protocol | Exceptional Human-AI Collaboration
An open-source framework for exceptional human-AI collaboration. Add genuine cognitive modes on top of plan/act tool-gating and break free of median-seeking...
|
76 % |
|
Foxtrot Jellyfish — Open-source agentic systems
Protocols, engines, and patterns for AI-augmented knowledge work. Platform agnostic. Open source. Built to outlast the tools.
|
76 % |
|
An open, methodology-first benchmark of AI app builders
BuilderProof is a community-editable wiki benchmarking AI app builders. Methodology-first, versioned and reproducible results.
|
76 % |
|
Verax | Independent AI test range
Verax is an independent test range where AI and autonomy systems are run against live adversaries and unscripted scenarios before they are fielded.
|
76 % |
|
Quesar by MLAI: Private AI operations
AI engineering engagements for traceable retrieval, bounded agent workflows, and private deployment.
|
76 % |
|
RedAssay — an AI pentest agent that proves what it finds
An AI pentest agent, run for you. It drives a real browser through your app like a user and proves every finding by deterministic replay. No third-party serv...
|
76 % |
|
Verdict Bridge — No model grades its own work
Adversarial multi-agent QA protocol for software development. Cross-vendor. No model grades its own work.
|
76 % |
|
Oculabs — One connected concussion recovery protocol, now AI-native
Oculabs turns fragmented concussion care into one system: a modular assessment library, configurable protocols, and AI built into every screen that summarize...
|
75 % |
|
Valhalla Protocol: Build the future.
Valhalla Protocol is an AI research and development company building intelligent systems, autonomous agents and AI-native products that turn ambition into re...
|
75 % |
|
Agora Protocol Hub: AI Agent Protocol Comparisons
Unbiased comparisons of AI agent protocols and multi-agent frameworks, starting with open-source Agora Protocol.
|
75 % |
|
Oraculum — The Compliance & Context Layer for Enterprise AI
Oraculum is the compliance and context layer for enterprise AI. Govern prompts, permissions, retrieval, model calls, and actions with full auditability.
|
75 % |
|
Web7 Foundation — The Accountability Layer for Agentic Systems
Web7 is the open accountability protocol for agentic systems: portable contracts for identity, delegation, policy, provenance, outcomes, and rollback.
|
75 % |
|
Reinstator · AI red teaming for applications and agents
Red-team your AI applications and tool-using agents across tool routes, indirect channels and persistent state. Every finding is confirmed in fresh sessions...
|
75 % |
|
PromptGolf - LeetCode for agentic prompting
Competitive benchmark for agentic prompting: fewer prompts, more passing tests.
|
75 % |
Detected social media networks