# Laminar

Aiolum story: https://aiolum.com/story/laminar/ · added 2026-11-12 · updated 2026-11-12

**In short.** Laminar (Observability for AI agents, United Kingdom): 3.5K GitHub stars, November 2026. Evidence grade E: no revenue figure was checked. Team: 2 founders; 6 people on the Y Combinator listing in November 2026. Launched September 2024; still running.

Observability for AI agents · United Kingdom · AI-native · team of 2 to 10 · still running · group: Developer tools and coding agents

- Outcome: 3.5K GitHub stars, November 2026 (evidence grade of the headline number: E; no revenue figure)
- Launch: September 2024; months from launch to the outcome: 26
- Team: 2 founders; 6 people on the Y Combinator listing in November 2026
- Outside funding: US$3M seed round, March 2026 (company blog)
- Main factor: Niche
- Meters, 1 to 4: evidence strength 1, copyable pattern 2, capital needed 2
- Tags: AI-native, Small team, Funded, Running, B2B, Developer tools

Links:

- Visit laminar.sh: https://laminar.sh/

**The product.** A platform that records what an AI agent did during a run and shows why it failed. It stores each model call and tool call, finds failure patterns across many runs, and lets an engineer rerun an agent from the failing step. The code is open source, and the company also hosts it.

**Why it worked.** Laminar serves one specific audience: teams whose AI agents run for a long time and fail in ways that are hard to repeat. The company wrote in March 2026 that older tools collect data and leave the engineer to find the cause. Its answer is a debugger, a failure detector and a search over all runs.

Evidence grades: A = disclosed by a public company, a filing or a payment processor, or verified by one; B = two or more independent reports agree, or a journalist confirmed it; C = a founder claim that no one else verified; D = reported, not verified: one founder claim or a third-party estimate with no second source; E = no revenue figure found, so the headline is a non-money fact and the other facts carry their own grades; calc = computed on this page from graded facts.

## Summary

Laminar is observability software for AI agents. Observability means recording what a program did, so that an engineer can find out why it failed afterwards. An AI agent is a program in which a language model decides the next step in a loop. Laminar records each step as a trace, the full list of calls in one run.

Robert Kim and Din Mailibay are listed by Y Combinator as the founders. The listing says the company was founded in 2024 and joined the Summer 2024 batch. Its first GitHub release is dated 5 September 2024. On 16 March 2026 the company wrote that it raised a US$3M seed round led by Atlantic.vc, and the listing gives a team of 6.

No revenue figure was found, so the story is grade E. Paid plans start at US$30 a month on the company's pricing page, and an account can be created without a sales call. No source states that a customer has paid, and the customers the company names are its own statement.

## Why did Laminar work?

### It narrowed the product to failures of long-running agents

The 2024 pages described a platform for pipelines, then observability for LLM apps. The March 2026 post speaks of agents that run for 40 minutes and fail in ways that are hard to reproduce. The home page lists a debugger, failure detection called Signals, evals and a SQL editor.

### The code is open source and can be self-hosted

The repository is licensed under Apache-2.0 and starts with Docker Compose. The company also hosts it. The company's March 2026 post says Laminar is fully open source. No source states how many customers came from the repository.

### The hosted plans charge for data volume, not for seats

Starter and Pro include unlimited seats and unlimited projects. Each plan includes an amount of data, and extra data costs US$2 or US$1.50 per GB, as the company's pricing page lists. The Signals analysis is billed apart, by tokens.

## How it was built

These details come from the company's own pages and repository. They are grade C.

- The server is written in Rust, according to the repository README and the company's blog posts.
- The first release used Postgres, RabbitMQ and ClickHouse. A later blog post says ClickHouse dictionaries give fast joins when a full message history is rebuilt at query time.
- The tracing library patches popular agent libraries with one line of code. The home page lists the Claude Agent SDK, the Vercel AI SDK, LiteLLM, Browser Use and others.
- Plans are paid by card through Stripe, as the privacy notice states.
- The company sells a second path for large buyers: on-premise deployment in the Enterprise plan, and Helm charts for Kubernetes.

Sources: [5, 4, 6, 14, 2, 7]

## Who tried this before Laminar, and why did they fail?

Three developer-tool companies in our database of failed startups tried to sell a better view of a developer's own code or terminal. The values below are from that database, were read on 12 November 2026 and are not independently verified.

The failed startups come from our database of dead startups, retold here in our own words; Its entries are AI-assisted summaries of public sources, so they carry no evidence grade.

### CodeSee (2020 to 2024; funding US$10.0M; cause: Monetization)

- Promise: Visual maps of a codebase that update when the code changes.
- Why it died: The record says engineering leaders did not see a clear return, and that such tools were cut first when budgets tightened.
- Record: 

### Fig DevTool (2020 to 2024; funding US$2.4M; cause: Monetization)

- Promise: Autocomplete and visual commands inside the terminal.
- Why it died: The record says free users did not become paying customers. It also says AWS acquired the company.
- Record: 

### Kite DevTool (2014 to 2024; funding US$21.0M; cause: Timing)

- Promise: AI code completion in the editor.
- Why it died: The record says it came seven years before GitHub Copilot and made monetization mistakes.
- Record: 

## What changed

The three records share one trouble: engineers liked the developer tool, but it was hard to sell. The records name free alternatives and a gain that was hard to measure. Laminar meets the same trouble in a new place. An AI agent that fails after 40 minutes costs real model spending and engineering time.

Two things differ from the records. The data Laminar stores comes from agents that call paid models, and the company prices by data volume. It also offers the code for free, so a team can try it before paying.

## Can one person copy Laminar now?

**Verdict: The open-source and per-GB structure can be copied; the result is not shown.**

A small team can publish the code of a developer tool, host it, and sell plans by data volume. The sources show the price list and a seed round, and they show no revenue. A reader should treat this as a pattern, not as proof that it earns.

1. Pick one failure that engineers of a new kind of software cannot yet debug.
2. Publish the code under a permissive licence, with a one-command self-hosting setup.
3. Offer a free hosted plan and two paid plans that scale with data volume.
4. Name the integrations on the home page, so that a team sees the setup is one line.

**What is still open.** Other long-running software that is hard to debug: voice agents, coding agents and data pipelines run by models.

## How much revenue does Laminar make? The numbers, graded

A funded team of 6 with an open-source product and a self-serve price list, and no revenue figure

| Fact | Value | Grade | Note | Sources |
|---|---|---|---|---|
| Company | LMNR AI, Inc., doing business as Laminar | C | Named in the company's terms of service, last updated 11 August 2026, which choose the law of Delaware. | [8] |
| Start | 2024, Y Combinator Summer 2024 batch | C | The Y Combinator listing gives Founded 2024, Batch Summer 2024 and Location London, United Kingdom. A page of the company from 4 March 2024 already shows the product, so the start is no later than March 2024. | [3, 10] |
| Founders and team | 2 founders; team size 6 | C | Y Combinator lists Robert Kim as founder and CEO and Din Mailibay as founder and CTO. The same listing says Team Size: 6. | [3] |
| First release of the open-source code | 5 September 2024, version v0.1.1-alpha | C | GitHub marks it as a pre-release. Its note reads Initial working version with docker compose -based self hosting, Postgres, RabbitMQ, and Clickhouse. | [6] |
| GitHub stars of the main repository | 3.5K stars, 262 forks, Apache-2.0 licence | C | Read on the repository page on 12 November 2026. | [5] |
| Prices | Free, Starter at US$30 and Pro at US$150 a month, Enterprise custom | C | The company's pricing page, read on 12 November 2026. | [1] |
| How a plan is bought | Get Started buttons lead to a sign-up page; card payment through Stripe | C | The sign-up page offers Google, GitHub and Microsoft accounts. The privacy notice says that if you purchase a paid plan, payment is handled by Stripe. | [1, 12, 7] |
| Funding | US$3M seed round, March 2026 | C | Company blog post of 16 March 2026 by Robert Kim. It names Atlantic.vc as lead, with Y Combinator and AAL.vc taking part. | [4] |
| Customers the company names | Browser Use, OpenHands, Rye.com, Alai and LegionIntel | C | The company's own blog post says teams at these companies rely on Laminar. No independent source was found. | [4] |
| Earlier product, March 2024 | A platform to build and deploy chains of model calls | C | The company page of 4 March 2024 offers building and monitoring pipelines of models. By 5 September 2024 the page speaks of observability for LLM apps. | [10, 11] |
| Browser agent Index | Open-source repository, 2.4K stars, Apache-2.0 | C | A repository of the same GitHub organization, archived on 3 July 2025. Its README describes a browser agent that runs web tasks and offers a serverless API. | [9] |
| Security claims | SOC 2 Type II and HIPAA | C | Stated on the company's home page and pricing page. No certificate was read. | [2, 1] |

How to read this table. Grade C means the company or a directory states the value. No value here is grade A or B. A calc value is not used.

## Pricing

Laminar prices read from the company's pricing page on 12 November 2026 (grade C)

| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Free |  | 0 | 1 GB data, 7 days of data access, 1 project, 1 seat, no Signals |
| Starter |  | 30 | 3 GB data, then US$2 per GB; 30 days of data access; unlimited projects and seats |
| Pro |  | 150 | 10 GB data, then US$1.50 per GB; 6 months of data access; unlimited projects and seats |
| Enterprise | custom | custom | custom limits, on-premise, dedicated support |

The page shows monthly prices only. Data above the included amount is billed per gigabyte. A separate Signals charge is billed by the tokens that Laminar's own model uses to analyze traces.

## When did Laminar launch, and how did it grow? Timeline

From an early platform for model pipelines in March 2024 to a hosted agent debugger with three paid tiers in November 2026.

- **4 March 2024** (grade C [10]): A copy of lmnr.ai in the Internet Archive offers a platform to build, deploy and monitor pipelines of models.
- **15 July 2024** (grade C [11]): A copy of the pricing page in the Internet Archive lists a Free plan and a Startup plan with a contact button.
- **Summer 2024** (grade C [3]): The company joins the Y Combinator Summer 2024 batch, according to the Y Combinator listing.
- **5 September 2024** (grade C [6, 10]): First release of the open-source code on GitHub. A copy of lmnr.ai of the same day speaks of observability for LLM apps.
- **16 March 2026** (grade C [4]): The company announces a US$3M seed round led by Atlantic.vc.
- **November 2026** (grade C [1, 13]): The pricing page lists Free, Starter and Pro plans, and the blog index shows a post dated 10 November 2026.

## What could go wrong for Laminar?

- **What is not verified.** No revenue figure was found, so the story is grade E. The pricing page, the home page, the blog and the repository were read; none gives revenue. A reader could look for a payment-processor profile or a press interview. No source states that a customer has paid, and the named customers are the company's own statement. The US$3M seed round is the company's own statement of 16 March 2026, and the team size of 6 is Y Combinator's listing.
- **Home-page trust signals are not sourced.** The home page shows one quotation from a person with the title CEO next to a Browser Use logo. It also shows SOC 2 Type II and HIPAA badges. A logo row names software libraries, not customers. No certificate or second source for the quotation was found.
- **Free and Starter workspaces may be used for training.** The terms say Privacy Mode is off by default on Free and Starter plans from 10 September 2026. While it is off, Laminar may use Signals run data from that workspace to train its Signals models, after redaction. It is on by default for Pro.
- **A sister browser agent was sold in 2025 and is archived.** The same GitHub organization published Index, an open-source browser agent, and archived the repository on 3 July 2025. The company's pricing page of 22 May 2025 sold Index agent steps at US$10 per 1,000 steps, as an Internet Archive copy shows. Today's pricing page does not list Index. The README and the archived documentation name no CAPTCHA, proxy or anti-bot feature. One source file of Index hides the browser's automation flag from websites, under a comment about avoiding detection as automation. The home page and the README name Browser Use and Stagehand as tools that Laminar can trace.

## Sources

1. Laminar pricing page — https://laminar.sh/pricing (grade C). Read in full on 12 November 2026. Source for the three paid and free plans and the Signals charge.
2. Laminar home page — https://laminar.sh/ (grade C). Read in full on 12 November 2026. Source for the product description, the integrations, the quotation and the security badges.
3. Y Combinator company page: Laminar — https://www.ycombinator.com/companies/laminar (grade C). Read in full on 12 November 2026. Source for the founders, the batch, the year, the location and the team size.
4. Laminar blog, 16 March 2026: Laminar raised $3M to build observability for long-running agents — https://laminar.sh/blog/2026-03-16-laminar-launch (grade C). Read in full on 12 November 2026. The company's own post. Source for the seed round and the named customers.
5. GitHub: lmnr-ai/lmnr — https://github.com/lmnr-ai/lmnr (grade C). Read the repository page on 12 November 2026. Source for stars, forks, licence and the feature list.
6. GitHub releases of lmnr-ai/lmnr, page 10 of 10 — https://github.com/lmnr-ai/lmnr/releases?page=10 (grade C). Read the oldest releases on 12 November 2026. Source for the first release on 5 September 2024 and its note.
7. Laminar privacy notice, updated 3 November 2026 — https://laminar.sh/policies/privacy (grade C). Read the sections on billing and the company name on 12 November 2026. Source for Stripe.
8. Laminar terms of service, updated 11 August 2026 — https://laminar.sh/policies/terms (grade C). Read in full on 12 November 2026. Source for the company name and the training licence.
9. GitHub: lmnr-ai/index — https://github.com/lmnr-ai/index (grade C). Read the repository page and the file index/browser/browser.py on 12 November 2026. Source for the browser agent Index and its archive date.
10. Internet Archive copies of lmnr.ai, 4 March 2024 and 5 September 2024 — https://web.archive.org/web/20240304005453/https://www.lmnr.ai/ (grade C). Read both copies in full. They show the earlier pipeline platform and the later observability page.
11. Internet Archive copy of lmnr.ai/pricing, 15 July 2024 — https://web.archive.org/web/20240715214436/https://www.lmnr.ai/pricing (grade C). Read in full. Source for the 2024 plans.
12. Laminar sign-up page — https://laminar.sh/sign-up (grade C). Read in full on 12 November 2026. It offers account creation with Google, GitHub or Microsoft.
13. Laminar blog index — https://laminar.sh/blog (grade C). Read in full on 12 November 2026. It lists posts from 3 December 2025 to 10 November 2026.
14. Laminar blog: How Laminar compresses agent traces by 20x — https://laminar.sh/blog/laminar-20x-agent-trace-compression (grade C). Read the section on storage on 12 November 2026. Source for ClickHouse.
15. Database of failed startups, local export — /method/#dead (grade C). Record pages read on the website on 12 November 2026. AI-written summaries, not independently verified.
16. Internet Archive copy of lmnr.ai/pricing, 22 May 2025 — https://web.archive.org/web/20250522070301/https://www.lmnr.ai/pricing (grade C). Read in full. Source for the Index agent steps sold in the 2025 plans.

Contact: aiolum@livyden.com. Run Laminar? Send corrections or a verified figure to aiolum@livyden.com (subject: Correction: Laminar).