Humanloop

Humanloop was an enterprise development platform for LLM applications, focused on evaluation, prompt management, and observability. Founded in 2020 by a team with backgrounds from Google Brain, Amazon Research, UCL, and Cambridge, the company positioned itself as one of the first dedicated platforms for managing and evaluating AI applications. The platform is now being sunset following an announcement that the Humanloop team is joining Anthropic.

What It Is

Humanloop provided a collaborative workspace where both engineers and non-technical team members — such as product managers and domain experts — could build, test, and monitor LLM-powered features. The platform covered three core areas: evaluation (understanding how AI systems perform), prompt management (versioning and deployment controls for prompts), and observability (monitoring and improving AI systems in production). It supported both UI-first and code-first workflows, enabling cross-functional teams to collaborate on AI product development without requiring every contributor to write code.

Core Platform Capabilities

The platform was organized around several functional pillars:

• Prompt Engineering: Collaborative workspace, multi-LLM playground, role-based access, prompt versioning, function calling, tagged deployments, and feedback collection.
• Evaluation: Eval reports, CI/CD integration, dataset versioning, offline and online evaluators, LLM-as-judge, human review, and code-first evaluation workflows.
• Observability: Online monitoring, distributed tracing, alerting, end-user feedback capture, and logging.
• Security & Compliance: SOC-2 Type II, GDPR, HIPAA (with BAAs), custom SSO + SAML, role-based access controls, VPC deployment, and EU or US hosting options.

Audience and Workflow

Humanloop served two primary user groups according to its own documentation: engineers who wanted to implement evaluations and monitoring in code, and product managers or domain experts who needed to work on prompt engineering and evaluation through a UI. The platform was designed to bridge these groups, allowing technical and non-technical contributors to collaborate in the same environment. Logs were created for each call to a Prompt, Tool, Evaluator, or Flow, capturing inputs, outputs, and metadata.

Update: Acquisition by Anthropic and Platform Sunset

The Humanloop homepage announces that the entire Humanloop team is joining Anthropic. The company describes this as a move to "amplify our impact" as the pace of AI progress accelerates. As part of this transition, the Humanloop platform is being sunset. The company has published a migration guide to help existing customers transition away from the platform. Humanloop was backed by Y Combinator, Index Ventures, Albion, Local Globe, UCLTF, and a number of angel investors. The founders — CEO Raza Habib (ML PhD, UCL), CPO Jordan Burgess (ML MPhil, Cambridge), and CTO Peter Hayes (ML PhD, UCL) — describe Humanloop as having been "the first development platform for LLM applications" and credit it with shaping "industry standards for how to manage and evaluate AI," though these are vendor-published claims.

Why It Matters

Humanloop's acquisition by Anthropic reflects the growing strategic importance of LLM evaluation and prompt management tooling. The platform addressed a real gap: before dedicated tools existed, teams relied on manual spreadsheets and ad-hoc processes for prompt iteration and model evaluation. Humanloop's approach — combining a collaborative UI with code-first APIs and enterprise-grade security — became a reference model for the LLMOps category. Its sunset marks the end of an independent product but signals that its capabilities and team will continue influencing AI development practices from within Anthropic.