AgentKit
AgentKit is a unified suite of tools designed to streamline the process of building, deploying, and optimizing AI agents. It introduces Agent Builder, a visual canvas that lets developers compose multi-agent workflows via drag-and-drop nodes, set guardrails, preview runs, and version workflows. The Connector Registry centralizes the management of data and tool integrations across workspaces and ensures governance and access control. ChatKit enables frictionless embedding of agentic chat interfaces, customizable to match branding and experience, into web or app environments. To support robust performance and reliability, AgentKit enhances its evaluation infrastructure with datasets, trace grading, automated prompt optimization, and support for third-party models. It also supports reinforcement fine-tuning to push agent capabilities further.
Learn more
Patronus AI
Patronus AI is an automated AI evaluation, security, and optimization platform for LLM applications and agentic systems. It helps teams confidently deploy AI products at scale by generating test suites, running experiments, logging traces, comparing outputs, monitoring production interactions, and evaluating model performance in real time. It provides industry-leading evaluators for RAG hallucinations, context quality, image relevance, answer correctness, prompt injection, sensitive data leakage, toxicity, bias, and other safety or reliability risks. Patronus Evaluators can score AI outputs on specific dimensions, and teams can also create custom evaluators for use-case-specific criteria. Its platform combines dashboards, APIs, plug-and-play evaluations, logs, traces, side-by-side comparisons, visualizations, analytics, and real-time alerts to help teams detect mistakes, benchmark models, improve prompts, and understand system behavior over time.
Learn more
Vivgrid
Vivgrid is a development platform for AI agents that emphasizes observability, debugging, safety, and global deployment infrastructure. It gives you full visibility into agent behavior, logging prompts, memory fetches, tool usage, and reasoning chains, letting developers trace where things break or deviate. You can test, evaluate, and enforce safety policies (like refusal rules or filters), and incorporate human-in-the-loop checks before going live. Vivgrid supports the orchestration of multi-agent systems with stateful memory, routing tasks dynamically across agent workflows. On the deployment side, it operates a globally distributed inference network to ensure low-latency (sub-50 ms) execution and exposes metrics like latency, cost, and usage in real time. It aims to simplify shipping resilient AI systems by combining debugging, evaluation, safety, and deployment into one stack, so you're not stitching together observability, infrastructure, and orchestration.
Learn more
Agenta
Agenta is an open-source LLMOps platform designed to help teams build reliable AI applications with integrated prompt management, evaluation workflows, and system observability. It centralizes all prompts, experiments, traces, and evaluations into one structured hub, eliminating scattered workflows across Slack, spreadsheets, and emails. With Agenta, teams can iterate on prompts collaboratively, compare models side-by-side, and maintain full version history for every change. Its evaluation tools replace guesswork with automated testing, LLM-as-a-judge, human annotation, and intermediate-step analysis. Observability features allow developers to trace failures, annotate logs, convert traces into tests, and monitor performance regressions in real time. Agenta helps AI teams transition from siloed experimentation to a unified, efficient LLMOps workflow for shipping more reliable agents and AI products.
Learn more
