Langfuse
Open-source tracing, evals and prompt management.
Acquired by ClickHouse (Jan 2026) - MIT core, self-hosting and Cloud continue · source
The LLMOps landscape by category - filterable by what actually drives the decision: open source, self-hostable, OpenTelemetry, compliance and team fit. Editorial, not paid placements.
Attributes last reviewed September 2026. Listings that carry a checked date were confirmed against the source page named on the card; the rest have not been re-checked and show no SOC 2 badge either way. *SOC 2 · cloud means the vendor's hosted service is certified - running the same tool yourself puts compliance back on you, which is why a self-hosted deployment never inherits the badge. Attributes change often - verify with the vendor before relying on them for procurement.
Maps to the Observability layer of the stack.
Open-source tracing, evals and prompt management.
Acquired by ClickHouse (Jan 2026) - MIT core, self-hosting and Cloud continue · source
Tracing and evals from the LangChain team.
Open-source GenAI tracing and evaluation, OpenTelemetry-compatible.
LLM tracing and evaluation. Source-available (Elastic License 2.0), free to self-host.
Proxy-based logging, caching and cost tracking.
Joined Mintlify (Mar 2026) - the service stays live in maintenance mode · source
Model monitoring and observability platform.
Maps to the Evaluation layer of the stack.
Eval and experimentation platform for LLM apps.
Datasets, LLM-as-judge and offline evals.
Open-source, pytest-style evaluation framework for LLM apps.
Sends usage telemetry by default - opt out with DEEPEVAL_TELEMETRY_OPT_OUT=1 · source
Open-source framework for model evaluations.
Open-source LLM testing library; the company now leads with AI red teaming.
Maps to the Prompt management layer of the stack.
Prompt registry, versioning and analytics.
Versioned prompts alongside traces and evals.
Acquired by ClickHouse (Jan 2026) - MIT core, self-hosting and Cloud continue · source
Prompt Registry with versions and aliases, next to tracing.
Maps to the RAG operations layer of the stack.
Managed vector database at scale.
Open-source vector database with hybrid search.
Open-source vector search engine.
Vector similarity search inside Postgres - no extra database to run.
Vector search on top of in-memory data.
Maps to the Security layer of the stack.
Open-source validation of LLM inputs and outputs.
NVIDIA's open-source toolkit for programmable input, output and dialog rails.
Prompt-injection and AI security guardrails.
Acquired by Check Point · source
Maps to the Deployment layer of the stack.
High-throughput open-source inference engine for serving LLMs.
Open-source model serving and packaging.
Serverless compute for AI workloads.
Run and deploy models via API.
Announced it is joining Cloudflare (Nov 2025) · source
Data + AI platform with model serving.
Maps to the Cost control layer of the stack.
Cost and token usage per trace and feature.
Acquired by ClickHouse (Jan 2026) - MIT core, self-hosting and Cloud continue · source
Open-source LLM gateway with per-key spend tracking, budgets and routing.
AI gateway with routing, caching and budgets.
No tools match all selected filters.
Missing a tool, or an attribute out of date?
This directory is community-maintained and vendor-neutral. Suggest a tool, or flag an attribute we should correct.
Contribute or correct →