The Agent Stack: Evaluation & Testing contains 12 companies across
2 categories.
Infrastructure used before production to test whether AI agents can complete intended tasks reliably. It covers simulated environments and functional evaluation. It excludes runtime security, policy enforcement, adversarial security testing, production control and audit evidence.
Simulated Environments & Test Worlds
Products that give agents realistic but non-production systems to act inside. They reproduce software state and workflows so teams can test agent behavior without touching live systems. General compute sandboxes and security red-teaming tools are excluded.
Era
Website: console.era.eon.io
Era is a simulated enterprise environment for developing and evaluating AI agents. It generates coherent synthetic companies across business systems and exposes them through vendor-compatible APIs and MCP servers.
Category: Simulated Environments & Test Worlds.
Funding stage: .
Founded: 2026.
Country: Unknown.
Landscape fit: It fits this category because Era provides realistic simulated company systems and synthetic data for agent evaluation without using live production systems.
Company timeline
-
2026-10-06:
Product snapshot
(product)
. Era generates synthetic company environments with connected records and history across emulated enterprise systems, for agent development and evaluation without real customer data.
Source: Era by Eon.
https://console.era.eon.io
-
2026-10-05:
Era launches with request access
(gtm)
. Eon launched Era as a free product and directed interested users to request access.
Source: Eon blog.
https://www.eon.io/blog/era-build-and-test-ai-agents
-
2026-10-05:
Era launches
(product)
. Eon introduced Era as a free tool for building and testing AI agents with simulated enterprise data and interfaces.
Source: Eon blog.
https://www.eon.io/blog/era-build-and-test-ai-agents
-
2026-09-24:
Hidden Knowledge benchmark published
(milestone)
. Era published a second enterprise-agent benchmark focused on questions where the correct answer must be inferred by connecting evidence across simulated company systems.
Source: arXiv.
https://arxiv.org/abs/2609.30055
-
2026-09-09:
Era benchmark published
(milestone)
. The Era by Eon Benchmark introduced a generated enterprise estate with exact ground truth for evaluating LLM agents across simulated business systems.
Source: arXiv.
https://arxiv.org/abs/2609.09853
Arga Labs
Website: argalabs.com
Arga Labs builds real world sandboxes that let teams test and train AI agents inside realistic replicas of external services before they touch production.
Category: Simulated Environments & Test Worlds.
Funding stage: Seed.
Founded: 2025.
Country: United States.
Employees: 2 to 10 employees.
Landscape fit: It fits Simulated Environments & Test Worlds because it provides realistic, resettable replicas of APIs and workflows so agents can be evaluated without affecting live systems.
Funding rounds
-
seed
announced 2026-08-26
: 10000000 USD
; investors: General Catalyst, Box Group, Emergence, Gradient, SV Angel
; source: TechCrunch on Arga Labs seed round
; https://techcrunch.com/2026/08/26/arga-is-building-a-better-way-to-train-enterprise-ai-agents
Company timeline
-
2026-10-06:
Arga publishes pricing
(gtm)
. Arga made its pricing public with tiered plans on its website.
Source: Arga pricing page.
https://www.argalabs.com/pricing
-
2026-10-06:
HR snapshot
(hr)
. The available evidence indicates a small company and no visible public hiring push. The careers page explicitly says there are no public openings right now.
Source: Arga Labs LinkedIn company page.
https://www.linkedin.com/company/arga-labs
-
2026-10-06:
Product snapshot
(product)
. Arga Labs provides sandboxes and stateful service twins for testing software and agents against external services. Its documentation describes browser-based test runs, repeatable scenarios, PR test runs, and APIs, CLIs, MCP interfaces, and SDKs for supported twins.
Source: Arga Labs home page.
https://www.argalabs.com
-
2026-10-01:
Arga shows active customer use
(gtm)
. Arga said customers were already using the product and had run more than 100,000 tests over the prior four months.
Source: Arga Labs LinkedIn company page.
https://www.linkedin.com/posts/arga-labs_today-arga-labs-is-excited-to-announce-our-activity-7498417231731544064-2aHp
-
2026-08-31:
Arga releases ArgaBench
(product)
. Arga published ArgaBench, a benchmark for evaluating multi app agents across real work tasks.
Source: ArgaBench: Where Frontier Agents Fail at Real-World Work | Arga Labs.
https://www.argalabs.com/blog/argabench-where-agents-fail
-
2026-08-26:
Seed
(funding)
. Amount: USD 10M. Investors: General Catalyst, Box Group, Emergence, Gradient, SV Angel
Source: TechCrunch on Arga Labs seed round.
https://techcrunch.com/2026/08/26/arga-is-building-a-better-way-to-train-enterprise-ai-agents
Waylabs
Website: waylabs.org
Waylabs builds synthetic enterprise environments that let teams train and test AI agents inside realistic ghost companies instead of production systems.
Category: Simulated Environments & Test Worlds.
Funding stage: Pre Seed.
Country: Turkey.
Employees: unknown.
Landscape fit: It fits this category because the product is a non production sandbox for evaluating agent behavior in realistic workflows, databases, APIs, and communication logs.
Company timeline
-
2026-10-06:
HR snapshot
(hr)
. The supplied Crunchbase summary lists a founder and CEO; the supplied evidence provides no verified employee count or clear public hiring signal.
Source: Crunchbase profile for WAY Labs.
https://www.crunchbase.com/organization/way-labs
-
2026-10-06:
Product snapshot
(product)
. Waylabs’ official site describes synthetic enterprise environments for AI agents, including generated company data, databases, APIs, communications, and business logic.
Source: Waylabs official website.
https://waylabs.org
Plato
Website: plato.so
Plato builds simulated workplaces that let AI labs train and evaluate agents on realistic job tasks without using live systems.
Category: Simulated Environments & Test Worlds.
Funding stage: Series A.
Founded: 2025.
Country: United States.
Employees: Not disclosed.
Landscape fit: Its simulated software and workflows give agents realistic, nonproduction environments for training and evaluation, matching this category. ([plato.so](https://plato.so/?utm_source=openai))
Funding rounds
-
Series A
announced 2026-10-28
: 16000000 USD
; investors: Nava Ventures, Gradient Ventures, Transpose Platform, Mustafa Suleyman
; source: PR Newswire
; https://www.prnewswire.com/news-releases/plato-raises-16-million-to-build-the-training-grounds-for-ais-next-leap-into-real-work-302921832.html
-
Seed
; investors: Gradient Ventures
; source: Plato / PR Newswire
; https://www.prnewswire.com/news-releases/plato-raises-16-million-to-build-the-training-grounds-for-ais-next-leap-into-real-work-302921832.html
-
Seed
; investors: Gradient Ventures
; source: Plato / PR Newswire
; https://www.prnewswire.com/news-releases/plato-raises-16-million-to-build-the-training-grounds-for-ais-next-leap-into-real-work-302921832.html
Company timeline
-
2026-11-01:
Five open roles listed on Plato’s careers page
(hr)
. Plato’s careers page listed five full-time openings across research, engineering, AI quality, and customer-facing work. This records the observed hiring status, not a dated hiring change.
Source: Plato.
https://plato.so/landing
-
2026-10-28:
Nava Ventures founder joined Plato’s board
(hr)
. Plato announced that Freddie Martignetti, founder of Nava Ventures, joined its board alongside the Series A announcement.
Source: PR Newswire.
https://www.prnewswire.com/news-releases/plato-raises-16-million-to-build-the-training-grounds-for-ais-next-leap-into-real-work-302921832.html
-
2026-10-28:
Series A
(funding)
. Amount: USD 16M. Investors: Nava Ventures, Gradient Ventures, Transpose Platform, Mustafa Suleyman
Source: PR Newswire.
https://www.prnewswire.com/news-releases/plato-raises-16-million-to-build-the-training-grounds-for-ais-next-leap-into-real-work-302921832.html
-
2026-10-27:
Plato released Computer Work Val v1.0
(product)
. Computer Work Val v1.0 is a 300-task environment data stream and validation set, spanning 210 simulated apps and usable for both testing and training.
Source: Plato.
https://plato.so/research/computer-work-val
-
2026-10-05:
Plato launched Web Clone Bench v1
(product)
. Web Clone Bench v1 uses nine reference web apps as tasks for models to create functional replicas, and benchmarks how well they reproduce those apps.
Source: Plato.
https://plato.so/research/web-clone-bench
-
Earlier seed financing led by Gradient Ventures
(funding)
. Plato disclosed an earlier seed financing led by Gradient Ventures in its October 28, 2026 Series A announcement. The amount and original closing date of the seed were not disclosed. The $16 million Series A brought total funding to $19 million, implying approximately $3 million of prior financing, but this is a calculation rather than a reported seed-round amount. This event deliberately has no occurrence date.
Source: Plato / PR Newswire.
https://www.prnewswire.com/news-releases/plato-raises-16-million-to-build-the-training-grounds-for-ais-next-leap-into-real-work-302921832.html
Agent Evaluation & Regression Testing
Products that measure whether an agent completes its intended tasks correctly and whether a new version improves or degrades performance. They focus on functional quality across full agent runs. Security testing and production monitoring are excluded.
Braintrust
Website: braintrust.dev
Braintrust provides an agent observability and evaluation platform that helps teams test AI agents, catch regressions, and improve quality before release.
Category: Agent Evaluation & Regression Testing.
Funding stage: Series B.
Founded: 2023.
Country: United States.
Employees: 51-200.
Landscape fit: It fits agent evaluation and regression testing because its core product is built to run evals, compare runs, and block quality regressions across full agent workflows.
Funding rounds
-
series_b
announced 2026-02-17
: 80000000 USD
; investors: ICONIQ, Andreessen Horowitz, Greylock, Elad Gil, basecase capital
; source: Braintrust's series B: building the infrastructure for production AI
; https://www.braintrust.dev/blog/announcing-series-b
-
series_a
announced 2024-10-08
: 36000000 USD
; investors: Andreessen Horowitz, Elad Gil, Greylock, Basecase, Datadog, Databricks Ventures, Guillermo Rauch, Simon Last, Bryan Helmig, Greg Brockman, Arthur Mensch
; source: Announcing our $36 million Series A
; https://www.braintrust.dev/blog/announcing-series-a
-
seed
announced 2023-12-13
: 5100000 USD
; investors: Greylock, Saam Motamedi
; source: Braintrust's seed round
; https://www.braintrust.dev/blog/seed-round
Company timeline
-
2026-10-06:
Braintrust posts broad hiring slate
(hr)
. Braintrust's careers page shows active openings across engineering, go to market, finance, legal, and support functions.
Source: Braintrust careers page.
https://www.braintrust.dev/careers
-
2026-10-06:
HR snapshot
(hr)
. The Braintrust careers page lists active openings across engineering, go to market, marketing, operations, sales, finance, legal, and support functions.
Source: Braintrust careers page.
https://www.braintrust.dev/careers
-
2026-10-06:
Product snapshot
(product)
. Braintrust is an agent observability and evaluation platform. It traces production calls, scores live traffic, supports datasets and experiments, and helps teams identify and investigate quality issues.
Source: AI observability platform page.
https://www.braintrust.dev/learn/ai-observability/v0
-
2026-09-03:
Braintrust adds Patterns and Debugger
(product)
. Braintrust announced Patterns for finding recurring issues and opportunities, Debugger for explaining likely failure modes in individual runs, and an enhanced Loop experience.
Source: One connected system for agent observability.
https://www.braintrust.dev/blog/active-observability-loop-patterns-debugger
-
2026-06-30:
Braintrust launches startup program
(gtm)
. Braintrust introduced a program through which approved startups can receive six to twelve months of Pro plan access, office hours, and invitations to company events.
Source: Agent observability for startups.
https://www.braintrust.dev/blog/braintrust-for-startups
-
2026-06-01:
Topics reaches general availability
(product)
. Braintrust said Topics was generally available and presented it as a foundation for active observability, continuously surfacing insights from production traces.
Source: AI observability is active observability.
https://www.braintrust.dev/blog/active-observability
LangChain
Website: langchain.com
LangChain builds software for developing and testing AI agents, including LangSmith for evaluation, debugging, monitoring, and testing of LLM and agent workflows.
Category: Agent Evaluation & Regression Testing.
Funding stage: Series B.
Founded: 2023.
Country: United States.
Employees: 201-500.
Landscape fit: It fits agent evaluation and regression testing because LangSmith is explicitly positioned for testing LLM applications, and LangChain says it added agent evaluation and related tooling for validating agent behavior across runs.
Funding rounds
-
series_b
announced 2025-10-20
: 125000000 USD
; investors: IVP, Sequoia, Benchmark, Amplify, CapitalG, Sapphire Ventures
; source: LangChain Series B
; https://www.langchain.com/blog/series-b
-
series_a
announced 2024-02-15
: 25000000 USD
; investors: Sequoia Capital
; source: LangSmith general availability and Series A
; https://www.langchain.com/blog/announcing-langsmith-ga
-
seed
announced 2023-04-04
: 10000000 USD
; investors: Benchmark
; source: LangChain seed round
; https://www.langchain.com/blog/announcing-our-10m-seed-round-led-by-benchmark
Company timeline
-
2026-10-06:
HR snapshot
(hr)
. The LinkedIn company page shows a 201–500 employee range and multiple active job listings across technical, support, and commercial roles. Its displayed count of 444 employees is a LinkedIn page figure.
Source: LangChain jobs on LinkedIn.
https://www.linkedin.com/company/langchain/jobs
-
2026-10-06:
Product snapshot
(product)
. LangChain provides open-source agent frameworks and describes LangSmith as its commercial agent engineering platform. The supplied evidence supports capabilities in building, controlling, testing, evaluating, observing, and improving agents.
Source: About LangChain.
https://www.langchain.com/about
-
2026-09-24:
LangSmith Engine v2 adds red teaming
(product)
. LangChain launched Engine v2 with proactive red teaming and automated validation of proposed fixes.
Source: New in LangSmith: red teaming and automated testing.
https://www.langchain.com/blog/langsmith-engine-v2-redteam
-
2026-09-17:
Deep Life Sci launches
(product)
. LangChain introduced Deep Life Sci, an open source agentic assistant for clinical and lab scientists.
Source: Building an Agent Harness for Life Sciences: Introducing Deep Life Sci.
https://www.langchain.com/blog/agent-harness-life-sciences
-
2026-09-01:
LangChain lists active job openings
(hr)
. LangChain's LinkedIn jobs page shows openings across analytics, technical support, enterprise sales, account management, and frontend platform engineering.
Source: LangChain jobs on LinkedIn.
https://www.linkedin.com/company/langchain/jobs
-
2026-06-01:
Rippling reports production use
(gtm)
. Rippling said its AI runs in production across millions of users globally on Deep Agents and LangSmith.
Source: How Rippling built production AI in 6 months with Deep Agents and LangSmith.
https://www.langchain.com/blog/how-rippling-went-ai-native-across-every-product-in-6-months-with-deep-agents-and-langsmith
Confident AI
Website: confident-ai.com
Confident AI is an AI quality platform for testing, evaluating, and improving AI applications and agents before and after deployment.
Category: Agent Evaluation & Regression Testing.
Funding stage: Seed.
Founded: 2023.
Country: United Kingdom.
Employees: 11-50.
Landscape fit: It fits this category because its core product measures whether AI systems, including agents, complete tasks correctly and whether changes improve or degrade quality across full runs. The company emphasizes evals in development, CI style regression tests, and result comparison rather than production monitoring or security testing alone.
Funding rounds
-
seed
announced 2025-03-12
: 2200000 USD
; investors: Y Combinator, Flex Capital, Oliver Jung, Vermilion Cliffs Ventures, Liquid 2 Ventures, January Capital, Rebel Fund
; source: Confident AI blog
; https://www.confident-ai.com/blog/how-i-closed-confident-ais-2-2m-seed-round-in-5-days
Company timeline
-
2026-10-06:
HR snapshot
(hr)
. Y Combinator lists Confident AI as a seven-person team. The YC jobs page shows four open roles, and the official careers page describes the company as a small, fast-moving team.
Source: Y Combinator.
https://www.ycombinator.com/companies/confident-ai/jobs
-
2026-10-06:
Product snapshot
(product)
. AI quality platform for testing, evaluating, monitoring, and improving AI applications and agents.
Source: Confident AI homepage.
https://www.confident-ai.com
-
2026-10-04:
Active roles listed
(hr)
. The YC jobs page lists four open positions at Confident AI.
Source: Y Combinator.
https://www.ycombinator.com/companies/confident-ai/jobs
-
2026-10-02:
Human feedback and observability updates shipped
(product)
. Confident AI shipped updates to annotation workflows, queue ingestion, validation, error analysis, and trace filtering.
Source: Confident AI docs changelog.
https://www.confident-ai.com/docs/changelog/2026/10/2
-
2026-09-24:
Customer Monitoring launched
(product)
. Confident AI launched Customer Monitoring to group users and traces by customer account.
Source: Confident AI blog.
https://www.confident-ai.com/blog/launch-week-q3-2026-day-4-customer-monitoring
-
2026-09-11:
Voice AI Evals launched
(product)
. Confident AI launched Voice AI Evals for testing the full voice conversation experience.
Source: Confident AI blog.
https://www.confident-ai.com/blog/launch-week-q3-2026-day-1-voice-ai-evals
Latitude
Website: latitude.so
Latitude is an open source platform for AI agent observability and evaluation. It traces agent runs, surfaces failures, and turns real production issues into evals that can catch regressions before users see them. ([latitude.so](https://latitude.so/?utm_source=openai))
Category: Agent Evaluation & Regression Testing.
Funding stage: Seed.
Founded: 2022.
Country: Spain.
Employees: 2-10.
Landscape fit: It fits Agent Evaluation & Regression Testing because the product focuses on converting production failures into evals and running those evals continuously to catch regressions across full agent runs. ([latitude.so](https://latitude.so/data-platform/evals))
Funding rounds
-
seed_vc
announced 2022-08-01
: 1320000 USD
; investors: Moonfire, Sur Ventures
; source: CB Insights company profile for Latitude
; https://www.cbinsights.com/company/latitude-364/financials
Company timeline
-
2026-10-06:
Latitude v1 shutdown announced
(product)
. Latitude's login page says v1 will shut down on October 31, 2026, and directs users to migrate to Latitude v2.
Source: Latitude login page.
https://app.latitude.so/login
-
2026-10-06:
HR snapshot
(hr)
. No company-specific employee count or recruiting activity is established by the supplied fetched evidence. The jobs.latitude.so page is a general AI job board whose visible listings are for other employers.
Source: Latitude AI engineering jobs board.
https://jobs.latitude.so
-
2026-10-06:
Product snapshot
(product)
. Latitude is an open-source observability and quality platform for AI agents. Its public product materials describe capturing production traces and tool calls, identifying recurring failures, and using signals, annotations, datasets, alerts, and coding-agent workflows to investigate and address issues.
Source: Latitude homepage.
https://latitude.so
-
2026-09-23:
Agent Score introduced
(product)
. Latitude introduced Agent Score, a rolling 0 to 100 score for agent quality. The score combines outcome, reliability, cost, speed, and safety.
Source: Latitude changelog.
https://latitude.so/changelog/agent-score
-
2026-09-04:
Saved searches become signals
(product)
. Latitude added the ability to create a signal directly from a saved search.
Source: Latitude changelog.
https://latitude.so/changelog
-
2026-08-14:
PII redaction added
(product)
. Latitude added span content filtering that removes emails, card numbers, secrets, and custom identifiers before storage.
Source: Latitude changelog.
https://latitude.so/changelog
Maxim AI
Website: getmaxim.ai
Maxim AI builds an enterprise AI evaluation, observability, and gateway platform that helps teams route, govern, test, and monitor production AI applications.
Category: Agent Evaluation & Regression Testing.
Funding stage: Seed.
Founded: 2023.
Country: United States.
Employees: 11-50.
Landscape fit: Maxim AI provides agent simulation and evaluation tooling for testing functional performance across scenarios and catching regressions before deployment.
Funding rounds
-
seed
announced 2024-06-18
: 3000000 USD
; investors: Elevation Capital, Undisclosed angel investors from Postman, Undisclosed angel investors from Chargebee, Undisclosed angel investors from Groww, Undisclosed angel investors from Razorpay, Undisclosed angel investors from Media.net
; source: Maxim AI official blog
; https://www.getmaxim.ai/blog/announcing-maxim-ais-general-availability-and-the-3m-funding-round-led-by-elevation-capital
Company timeline
-
2026-09-07:
HR snapshot
(hr)
. Maxim AI appears to be a small enterprise software team with limited public hiring. The company is led by its two founders and is recruiting selectively rather than running a broad hiring push.
Source: LinkedIn company page for Bifrost by Maxim AI.
https://www.linkedin.com/company/bifrost-ai-gateway
-
2026-09-07:
Product snapshot
(product)
. Maxim AI is an end to end platform for simulation, evaluation, observability, and governance of AI agents and applications. Its public product surface also includes Bifrost, an enterprise AI gateway for routing, guardrails, and traffic control.
Source: Platform Overview.
https://www.getmaxim.ai/docs/introduction/overview
-
2026-01-16:
Maxim updates product
(product)
. Maxim’s changelog highlights a logging and observability overhaul, an MCP gateway, and evals on file attachments.
Source: Maxim Updates - Maxim Blog.
https://www.getmaxim.ai/blog/tag/maxim-updates
-
2025-07-04:
Bifrost publicly released
(product)
. Maxim AI announced the public release of Bifrost, its open-source LLM gateway for high-throughput production AI systems, with routing, extensibility, and built-in observability.
Source: Maxim AI June 2025 updates.
https://www.getmaxim.ai/blog/maxim-ai-june-2025-updates/
-
2025-05-12:
Maxim publishes Clinc story
(gtm)
. Maxim said Clinc used its platform to benchmark and refine a conversational AI system for banking.
Source: Clinc customer story.
https://www.getmaxim.ai/blog/elevating-conversational-banking-clincs-path-to-ai-confidence-with-maxim
-
2025-05-01:
Maxim publishes Atomicwork story
(gtm)
. Maxim said Atomicwork used its platform to simulate user interactions before release and to support regression testing and validation of agent workflows.
Source: Atomicwork customer story.
https://www.getmaxim.ai/blog/scaling-enterprise-support-atomicworks-journey-to-seamless-ai-quality-with-maxim
Patronus AI
Website: patronus.ai
Patronus AI builds evaluation and testing software that helps teams measure whether AI systems and agents work correctly and safely, then catch regressions before deployment.
Category: Agent Evaluation & Regression Testing.
Funding stage: Series B.
Founded: 2023.
Country: United States.
Employees: 51-200.
Landscape fit: It fits agent evaluation and regression testing because its core product measures model and agent performance, generates adversarial test cases, and uses benchmarks and simulation to detect failures across AI runs.
Funding rounds
-
series_b
announced 2026-06-25
: 50000000 USD
; investors: Greenfield Partners, Lightspeed Venture Partners, Notable Capital, Datadog, Samsung, Factorial Capital, Gokul Rajaram, others
; source: Patronus AI raises $50 million Series B
; https://www.patronus.ai/announcements/announcing-our-50m-series-b
-
investment
announced 2024-10-15
; investors: InvestInData
; source: Announcing an Investment from InvestInData
; https://www.patronus.ai/announcements/announcing-an-investment-from-investindata
-
series_a
announced 2024-05-22
: 17000000 USD
; investors: Notable Capital, Lightspeed Venture Partners, Datadog, Gokul Rajaram, Factorial Capital, Jonathan Frankle, Jason Warner, Tristan Handy, Michael Callahan, Barr Moses, Aparna Sinha, Nadim Hossain, others
; source: Announcing our $17M Series A
; https://www.patronus.ai/blog/announcing-our-17-million-series-a
-
seed
announced 2023-09-01
: 3000000 USD
; investors: Lightspeed Venture Partners, Factorial Capital, Amjad Masad, Gokul Rajaram, Michael Callahan, Prasanna Gopalakrishnan, Suja Chandrasekaran, others
; source: Patronus AI launches out of stealth
; https://patronus.ai/blog/patronus-launch
Company timeline
-
2026-10-06:
HR snapshot
(hr)
. The Greenhouse board lists eight open roles spanning operations, finance, research, and engineering. Patronus AI's join page describes its team as AI researchers and engineers in San Francisco.
Source: Patronus AI jobs board.
https://job-boards.greenhouse.io/patronusaiinc?from_theconsensus=1
-
2026-10-06:
Product snapshot
(product)
. Patronus AI provides AI evaluation and guardrail tools, including evaluator models and APIs, and describes newer multimodal evaluation and digital-world simulation work.
Source: Patronus AI launches out of stealth.
https://patronus.ai/blog/patronus-launch
-
2026-10-01:
Portkey integration announced
(gtm)
. Patronus AI announced an integration with Portkey that gives Portkey users access to Patronus guardrail models.
Source: Patronus AI and Portkey integration.
https://www.patronus.ai/announcements/introducing-the-patronus-portkey-integration
-
2026-09-03:
SpeedrunBench introduced
(product)
. Patronus introduced SpeedrunBench, a benchmark for frontier agents that measures how efficiently agents complete long-horizon game tasks rather than only whether they succeed.
Source: Patronus AI.
https://www.patronus.ai/blog/%20introducing-speedrunbench-a-challenging-benchmark-for-frontier-agents
-
2026-09-01:
Eight open roles listed on jobs board
(hr)
. Patronus AI's Greenhouse board lists eight open roles across operations, finance, research, and engineering.
Source: Patronus AI jobs board.
https://job-boards.greenhouse.io/patronusaiinc?from_theconsensus=1
-
2026-08-20:
FigmaTrace released
(product)
. Patronus released FigmaTrace, a dataset of 3,469 Figma design trajectories covering more than 200 hours of human workflow data for training and evaluating computer-use agents.
Source: Patronus AI.
https://www.patronus.ai/blog/figmatrace-a-comprehensive-training-dataset-for-figma-design-workflows
Langfuse
Website: langfuse.com
Langfuse is an open source AI engineering platform that helps teams trace, evaluate, and improve LLM applications and AI agents in production.
Category: Agent Evaluation & Regression Testing.
Funding stage: Acquired.
Founded: 2023.
Country: Germany.
Employees: 16+.
Landscape fit: It fits agent evaluation and regression testing because it provides datasets, LLM as judge evaluations, and production tracing to measure whether agent runs work correctly and whether changes improve or degrade results.
Funding rounds
-
seed
announced 2023-11-07
: 4000000 USD
; investors: Lightspeed Venture Partners, La Famiglia, Y Combinator
; source: Langfuse
; https://langfuse.com/blog/announcing-our-seed-round
Company timeline
-
2026-10-06:
HR snapshot
(hr)
. Langfuse's official pages list 16+ full-time employees and describe current recruiting across company functions. The company identifies Berlin and San Francisco as offices and says European product and engineering roles are remote-first.
Source: Careers at Langfuse.
https://langfuse.com/careers
-
2026-10-06:
Product snapshot
(product)
. Langfuse is an open-source AI engineering platform for developing, monitoring, evaluating, and debugging AI applications. It provides observability, prompt management, evaluations, datasets, experiments, a playground, APIs, and deployment options including cloud and self-hosting.
Source: Langfuse.
https://langfuse.com
-
2026-10-05:
Jev evaluators ship
(product)
. Langfuse added support for using TypeSafe's Jev decision model as a judge in evaluation workflows.
Source: Langfuse GitHub releases.
https://github.com/langfuse/langfuse/releases
-
2026-09-01:
Multi modal evaluators ship
(product)
. Langfuse added evaluator support for images, audio, video, PDFs, and text files captured in traces.
Source: Langfuse Changelog.
https://langfuse.com/changelog
-
2026-08-31:
Langfuse v4 goes live
(product)
. Langfuse announced v4 as live on Cloud and generally available for self hosted deployments.
Source: Langfuse August Update.
https://langfuse.com/blog/2026-08-31-langfuse-august-update
-
2026-08-21:
CLI 1.0 ships
(product)
. Langfuse released CLI 1.0 with a rewritten command line tool and faster startup.
Source: Langfuse Changelog.
https://langfuse.com/changelog
Crux Labs
Website: crux-labs.ai
Crux Labs turns expert judgment into custom evaluations that test AI agents on real tasks and track whether their performance changes.
Category: Agent Evaluation & Regression Testing.
Country: United States.
Employees: Not disclosed.
Landscape fit: It creates custom evaluations from expert judgment and replays agent traces to check whether updates improve or degrade task performance. Its production monitoring features extend beyond the category, but the core product fits agent evaluation and regression testing.
Funding rounds
-
Undisclosed
announced 2026-10-29
: 6000000 USD
; investors: Benchstrength, Misneach, Analog Ventures, Pareto Holdings
; source: Crux Labs / PR Newswire
; https://www.prnewswire.com/news-releases/crux-labs-raises-over-6m-to-align-ai-agents-with-human-judgment-inside-the-enterprise-302923216.html
Company timeline
-
2026-10-29:
Announced more than $6M in financing
(funding)
. Crux Labs announced an oversubscribed financing of more than $6 million, co-led by Benchstrength and Misneach, with Analog Ventures and Pareto Holdings participating. The company did not disclose the round type. The amount is a lower bound, not an exact round total.
Source: Crux Labs / PR Newswire.
https://www.prnewswire.com/news-releases/crux-labs-raises-over-6m-to-align-ai-agents-with-human-judgment-inside-the-enterprise-302923216.html
-
2026-10-29:
Crux Labs emerges from stealth and introduces its platform
(product)
. Crux Labs publicly introduced its platform for capturing expert judgment, creating and calibrating agent evaluations, monitoring production traffic, and improving agents.
Source: PR Newswire.
https://www.prnewswire.com/news-releases/crux-labs-raises-over-6m-to-align-ai-agents-with-human-judgment-inside-the-enterprise-302923216.html
-
2026-10-01:
Crux reports enterprise deployments
(gtm)
. Crux Labs reported traction and deployments with high-growth startups and Fortune 500 enterprises across healthcare, cybersecurity, financial services, and other complex industries. It did not name customers or quantify deployment scale.
Source: PR Newswire.
https://www.prnewswire.com/news-releases/crux-labs-raises-over-6m-to-align-ai-agents-with-human-judgment-inside-the-enterprise-302923216.html
