Vectorize (Hindsight Cloud): Product, team and how it was built
Vectorize (Hindsight Cloud) (Open-source memory for AI agents, with a hosted pay-as-you-go cloud, United States): 60.6K GitHub stars and a pay-as-you-go cloud, November 2026. Evidence grade E: no revenue figure was checked. Team: 2 engineers at founding in January 2024 (about page); 7 employees in October 2025 (Latka, from LinkedIn); no newer count found. Launched January 2024; still running.Added
- Outcome
- 60.6K GitHub stars and a pay-as-you-go cloud, November 2026
- Evidence
- E no revenue figure was checked
- Launch
- January 2024
- Status
- Still running
- Team
- 2 engineers at founding in January 2024 (about page); 7 employees in October 2025 (Latka, from LinkedIn); no newer count found
- Outside funding
- US$3.6M seed round led by True Ventures, April 2024 (the company's about page)
- Main channel
- Open source
What is not verified
No revenue figure and no customer count was found. Latka estimates about US$1M a year in October 2025 and says the company has not published its revenue. A payment-processor profile, a press article or a founder interview could give a figure. The company states the US$3.6M seed round on its own about page, and no press article about it was read. The benchmark claims are the company's own. No source states that a customer has paid. This story is grade E: the business is real and its product, price and customers are described from public pages, but no revenue figure was checked. Aiolum shows it as an idea worth studying, not as proof that it makes money.
Vectorize sells Hindsight, an agent memory system. An agent memory system stores what an AI agent learns from users and tasks, so a later session can use it. The code is open source under the MIT licence. Hindsight Cloud is the hosted version that the company runs and bills by use.
The company's about page says two engineers founded it in January 2024. It says a US$3.6M seed round led by True Ventures closed in April 2024. The first product was a data platform for RAG, retrieval-augmented generation, which opened to the public on 14 May 2024, the CEO wrote. GitHub lists the first Hindsight release on 4 December 2025.
No revenue figure was found, so the story is grade E. The pricing page lists self-service Hindsight Cloud with a US$5 free credit, and the billing documentation says credits are bought through Stripe Checkout. No source states that a customer has paid.

Why did Vectorize (Hindsight Cloud) work?
Four causes, in the order they mattered. Each one is a reusable lesson.
The code is free and public, and the cloud is the paid layer
The pricing page says Hindsight is MIT licensed with no usage limits, and that Hindsight Cloud is the managed version. GitHub shows 60.6K stars and 7.4K forks. The home page has the buttons "View on GitHub" and "Start free". The sources do not say how many Cloud customers came from GitHub.
The hosted version is billed by use with no monthly fee
The pricing page says Cloud has no fixed monthly fee and no seat price. Each operation has its own unit. The billing documentation says consolidation is included in the retain price. It also warns that automatic refresh of knowledge pages can make that line grow with write volume.
Enterprise and consulting sit beside the self-service plan
The pricing page sends Enterprise buyers to sales. They get a private cloud, single sign-on, audit logs and invoiced billing. The home page also offers implementation services, team training and architecture consulting. The sources give no figures for either.
How it was built
These details come from the project's GitHub page, the company's pricing page and its documentation. They are grade B or C. [7, 2, 4]
- Hindsight has three operations: retain stores memories, recall retrieves them and reflect reasons over them. The README names them.
- Storage is PostgreSQL with pgvector, or Oracle AI Database 23ai, the README says. The pricing page says the self-hosted version uses embedded PostgreSQL.
- The server starts with one Docker command and works with more than 25 model providers, the README says. Clients exist for Python, TypeScript and Go.
- The company trains its own small models for decisions, reranking and embeddings. The pricing page gives open weights for them on Hugging Face.
- Billing for Cloud uses prepaid credits through Stripe Checkout, the billing documentation says. The documentation site is built with Docusaurus.
Tools named in the sources
Each one opens its page in the Tools tab, with every story that uses it.
Who tried this before Vectorize (Hindsight Cloud), and why did they fail?
Four companies in our database of failed startups sold developer infrastructure or tools in markets with strong free alternatives. The values below are from that database and are not independently verified.
RethinkDB
The promiseAn open-source database that pushed live query results to applications.
Why it diedThe database says developers liked the idea but did not pay for it. The only paid paths were support and hosting, and each customer needed much help.
Kite DevTool
The promiseAI code completion inside the editors that developers already used.
Why it diedThe database names the launch of GitHub Copilot in June 2021 as the primary cause. It says paid conversions fell.
Fig DevTool
The promiseAutocomplete and visual command builders for the terminal.
Why it diedThe database says users loved it but few converted to paid plans. It says the free tier was too good.
DotCloud
The promiseA platform that let developers deploy applications without managing servers.
Why it diedThe database says its multi-language platform was costly to run. It says Docker, built inside the company, made the platform itself obsolete.
The failed startups come from our database of 1,749 dead startups, retold here in our own words. Its entries are AI-assisted summaries of public sources, so these cards carry no evidence grade.
What changed
Each of the four sold to developers who had a free or cheaper way to do the same job. RethinkDB and Fig struggled to turn free users into paying ones. Kite lost its market to a free tool that came with the editor. DotCloud found that its own spin-off was the better product.
Hindsight starts from the free side on purpose. The MIT code runs anywhere, and the company charges for hosting, scaling and enterprise controls. The sources do not show that this has solved the payment problem. They show a published price list, a self-service credit system and a large star count.
Can one person copy Vectorize (Hindsight Cloud) now?
The structure can be copied. The paying demand is not shown.
A free open-source engine with a hosted version billed by use is a structure that a small funded team can run. The sources show wide interest in the free code. They do not show how many people pay for the hosted version.
Where it is still open: Other shared needs of agents: access control for shared memory, audit trails, evaluation of agent behaviour and tools that run for hours.
- 1Pick a gap that AI agents share, such as forgetting everything between sessions.
- 2Release the engine under a permissive licence and make it start with one command.
- 3Sell the parts a customer does not want to run: hosting, scaling, backups and security controls.
- 4Publish unit prices and a free starting credit, so a reader sees what each operation costs.
How much revenue does Vectorize (Hindsight Cloud) make?
Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.
| Fact | Value | Grade | How we know | Sources |
|---|---|---|---|---|
| Start of the company | January 2024 | C | The about page timeline says "Founded company" in January 2024 and "Two engineers set out to solve how AI agents store and learn from experience." An archived about page of 5 February 2024 says the founders "started building Vectorize in 2023". Latka also says 2023. | [1, 11, 15] |
| Named leaders | Chris Latimer, co-founder and CEO; Chris Bartholomew, co-founder and CTO | C | The about page lists both under "Meet our leadership". An archived about page of 5 February 2024 lists both under "Meet our founders" as co-founders. | [1, 15] |
| Team size | 7 employees, October 2025 | C | Latka gives the count and says it comes from LinkedIn. The careers page lists open roles and no headcount. | [11, 12] |
| Seed round | US$3.6M, April 2024, led by True Ventures | C | The company's about page states the amount. The True Ventures companies page lists Vectorize with a date of 4 April 2024 and no amount. | [1, 10] |
| First product opened to the public | 14 May 2024 | C | Blog post by the CEO. A Wayback copy of the home page from 5 February 2024 already shows the company's data platform pitch. | [5, 6] |
| First Hindsight release on GitHub | 4 December 2025, version 0.0.16 | B | The last page of the releases list. A blog post by the CEO introduces Hindsight on 16 December 2025. | [8, 9] |
| GitHub stars, forks and commits | 60.6K stars, 7.4K forks, 4,120 commits | B | The repository page, read 12 November 2026. The licence shown is MIT. | [7] |
| Latest release | version 1.2.1, 10 November 2026 | B | The first page of the GitHub releases list. | [8] |
| Hindsight Cloud usage prices | retain US$10 and recall US$0.75 per 1M tokens | C | The company's pricing page. The billing documentation lists the same rates and says the unit for recall is output tokens. | [2, 4] |
| How a customer pays | self-service credits through Stripe Checkout | C | The billing documentation says owners and admins can purchase credits and manage payment methods. Its payment section names Stripe Checkout. | [4] |
| Benchmark claim | 94.6% on LongMemEval-S against 74.0% for the next best system | C | The company's own claim on its home and about pages. No independent test of it was found. | [3, 1] |
| Startup programme | three credit steps for startups | C | The startups page lists US$1,000 of credits, then US$4,000 and US$10,000 of bonus credits, as the company states. | [13] |
How to read this table. A grade C value comes from the company's own pages or a third-party page that says so. A grade B value is a timestamp or count kept by GitHub. No value is a money figure of revenue.
| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Self-hosted, MIT licence | 0 | 0 | your own servers; community support |
| Hindsight Cloud, retain | usage | US$10 per 1M tokens stored as memories | |
| Hindsight Cloud, recall | usage | US$0.75 per 1M tokens retrieved | |
| Hindsight Cloud, storage | usage | US$0.25 per 1M tokens a month, after 30 days | |
| Enterprise | contact | contact | private cloud, own cloud account or on premises |
Hindsight Cloud has no fixed monthly fee and no seat price. The company's pricing page says new accounts start with a US$5 credit and lists the unit prices of each operation. Reflect and mental-model refresh cost US$0.05 per call. Enterprise has no public price.
When did Vectorize (Hindsight Cloud) launch, and how did it grow?
From a retrieval data platform in 2024 to open-source agent memory and a pay-as-you-go cloud in 2025 and 2026. Each entry carries its grade.
- January 2024CThe about page dates the founding of the company to this month. A Wayback copy of 5 February 2024 shows the home page "Connect your data to your LLMs". [1, 6]
- April 2024CThe about page says a US$3.6M seed round led by True Ventures closed. True Ventures lists Vectorize with the date 4 April 2024. [1, 10]
- 14 May 2024CThe CEO announces that the Vectorize platform for building RAG applications is publicly available and free to try. [5]
- 4 December 2025BThe first Hindsight release on the GitHub list, version 0.0.16, is published. The CEO introduces Hindsight in a blog post on 16 December 2025. [8, 9]
- 12 November 2026CThe pricing page lists self-hosted, Cloud and Enterprise. GitHub shows 60.6K stars, and version 1.2.1 is two days old. [2, 7, 8]
What could go wrong for Vectorize (Hindsight Cloud)?
The risks are the mirror image of the causes above.
What is not verified
No revenue figure and no customer count was found. Latka estimates about US$1M a year in October 2025 and says the company has not published its revenue. A payment-processor profile, a press article or a founder interview could give a figure. The company states the US$3.6M seed round on its own about page, and no press article about it was read. The benchmark claims are the company's own. No source states that a customer has paid.
Counters, logos and badges
The home page shows the heading "A few of the teams building with Vectorize" and logos of Brightness, Bronson, Electronic Arts, Groq and Nvidia. No page says what these companies use or pay. The about page shows "15,000+ Developers", and the pages show SOC 2 badges. No testimonial or named quotation was found. No source dates or explains these counters.
Start date, earlier product and team
The company dates its founding to January 2024. Its archived about page of February 2024 says the founders started building in 2023, and Latka also says 2023. The first product was a data platform for RAG applications, before Hindsight. The True Ventures page draws a line from Sailplane to Vectorize dated 2024. That page says each line is "a founder who brought in another". Sailplane is a separate company in the same list, not an earlier name. GitHub shows that the organisation account vectorize-io was created on 3 April 2021. Its oldest public repository is from January 2024, and no source explains the earlier date. The latest headcount is from October 2025, and the careers page lists many open roles.
Other offers
The home page offers implementation services, team training and architecture consulting next to the product. The terms and privacy pages load inside an embedded viewer and could not be read. The billing documentation warns that automatic refresh of knowledge pages can make a bill larger than expected. The README says Hindsight can also run on a user's ChatGPT, Claude, Cursor or GitHub Copilot subscription with no API key. No page read says whether those providers permit this use.
Sources
The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.
- C Vectorize about page, read 12 November 2026Read in full. Source for the founding date, the seed round, the two named leaders and the 15,000 developers counter.
- C Vectorize pricing page, read 12 November 2026Read in full. Source for the tiers, the unit prices, the free credit and the FAQ.
- C Vectorize home page, read 12 November 2026Read in full. Source for the product description, the benchmark claim, the logo wall and the consulting offer.
- C Hindsight Cloud billing documentation, read 12 November 2026Read the first part and the pricing table. Source for self-service credits, Stripe Checkout and the rates.
- C Chris Latimer, 14 May 2024: Introducing Vectorize: The easy path to accurate RAG applicationsRead in full. Source for the public availability of the first product.
- C Wayback copy of vectorize.io, 5 February 2024Read the home page text. Source for the early product pitch and its date.
- B GitHub repository vectorize-io/hindsight, read 12 November 2026Read the main page and the README. Source for stars, forks, commits, the licence and the architecture. The GitHub API gave 60,558 stars and the creation date of the organisation account.
- B GitHub releases of vectorize-io/hindsight, first and last page, read 12 November 2026Read the first page and page 8 of the list. Source for the latest release and the earliest release entry.
- C Chris Latimer, 16 December 2025: Introducing Hindsight: Agent Memory That Works Like Human MemoryRead the opening and the benchmark section. Source for the date of the Hindsight introduction.
- C True Ventures companies page, read 12 November 2026Read the page text and its embedded data. Source for the listing of Vectorize and the date 4 April 2024.
- D Latka company page for Vectorize, read 12 November 2026Read in full. A third-party estimate. Source for the employee count and the revenue estimate that the page says is unpublished.
- C Vectorize careers page, read 12 November 2026Read in full. Source for the open roles and the absence of a headcount.
- C Vectorize startups page, read 12 November 2026Read in full. Source for the credits programme.
- C Database of failed startups, local exportRecord pages read on the website on the day of writing. AI-written summaries, not independently verified. Source for the failed predecessors.
- C Wayback copy of the Vectorize about page, 5 February 2024Read in full. Source for the two co-founders and for the start of work in 2023.
Run this company? Send corrections or a verified figure to aiolum@livyden.com