The promise of autonomous AI agents is often framed as a problem of intelligence. The more immediate, practical bottleneck is memory. For an agent to reason about a user's behavior, a device's state, or its own past actions, it needs to query a complete, contextual history of events. That is a task for which traditional tabular databases were never designed, creating a gap that a new class of infrastructure is beginning to target.
Keewano, a Tel Aviv-based startup founded in 2024, is making its move into that gap. In September 2026, the company publicly launched KeewanoDB, an event-oriented database built specifically for AI agents and machine reasoning, alongside a $12 million seed round [PR Newswire, September 2026]. The core bet is that storing actions as ordered, contextual sequences will let agents analyze behavioral histories without the costly reconstruction required from conventional tables [Hetz Ventures, Unknown]. For engineering teams building agentic systems, the proposed payoff is speed. The company claims its system can query 250 million events in less than half a second [Citybiz, Unknown].
The Wedge: From Human Queries to Agent Reasoning
The fundamental limitation Keewano aims to solve is one of data shape. A relational database is optimized for human analysts running predefined SQL queries against structured tables. An AI agent, however, needs to investigate unfamiliar patterns across long, complex event histories. It needs to ask open-ended questions like "what changed before this user churned?" or "which sequence of actions typically leads to a system failure?"
KeewanoDB approaches this by treating every action from users, devices, or other agents as a discrete event stored in its complete context. This creates a timeline that an agent can traverse and interrogate directly. The company has built a suite of tools on top of this core engine, including Keewano Analytics for open-ended questioning and a Slack integration called Keewano Ask [Keewano, retrieved 2026]. Perhaps most tellingly for its target market, the company also offers Keewano MCP for Agents, a model context protocol server that gives AI agents direct, programmatic access to the analytics layer [Keewano, retrieved 2026].
A Team Built on Gaming-Scale Analytics
The ambition to handle hundreds of millions of events at sub-second latency does not emerge from a vacuum. It comes from a founding team with a decade of experience running analytics at the scale of massive multiplayer games. The four co-founders, Mark Kardashov (CEO), Dima Karger, Pavel Bibergal (CTO), and Vitaly Bukhovsky, collectively have two prior exits to their names [PR Newswire, September 2026].
Their backgrounds point directly to the performance demands of Keewano's target market. Pavel Bibergal was previously CTO at gaming giant Plarium, and Dima Karger led one of its largest studios, scaling games to millions of concurrent players [Gamigion, Unknown]. This experience in building and leading massive engineering organizations for resilient, high-throughput systems is the substrate for Keewano's technical architecture. The team's prior ventures, TestProject (acquired by Tricentis) and Devalore (acquired by Abra), further cement their operator credentials [FundUp, Unknown].
| Founder | Role at Keewano | Key Prior Experience |
|---|---|---|
| Mark Kardashov | Co-Founder, CEO | Co-founded TestProject (acquired by Tricentis) [PR Newswire, September 2026] |
| Dima Karger | Co-Founder | Led a major studio at Plarium, scaling to millions of players [Gamigion, Unknown] |
| Pavel Bibergal | Co-Founder, CTO | Former CTO at Plarium [Gamigion, Unknown] |
| Vitaly Bukhovsky | Co-Founder, CISO | Over 20 years in cybersecurity, with prior exits as a CTO [a16z Speedrun, Unknown] |
The Seed Round and Strategic Position
The $12 million seed round, led by Hetz Ventures with participation from a16z speedrun, Remagine Ventures, and DIG Ventures, arrived as the product launched [PR Newswire, September 2026]. The capital is presumably earmarked for scaling the engineering team and developer outreach, a signal reinforced by a job posting for a Head of Developer Relations spotted in July 2026 [Hetz Ventures Careers, July 2026].
A notable aspect of the company's stated model is its pricing philosophy. According to lead investor Hetz Ventures, KeewanoDB does not charge per event [Hetz Ventures, Unknown]. In a market where cloud database costs can spiral with data volume, this could be a significant point of differentiation for cost-conscious engineering teams planning to store vast event histories.
The company offers both a local version, KeewanoDB Local, for development and smaller workloads, and a managed cloud service, KeewanoDB Cloud, for distributed production deployments [Keewano, retrieved 2026]. This hybrid approach caters to developers who want to start locally and enterprises that need scalable, managed infrastructure.
The Roadmap and Inherent Risks
For all its technical promise and experienced backing, Keewano is entering a foundational software market dominated by entrenched incumbents like PostgreSQL, MongoDB, and specialized time-series databases. Its success hinges on convincing developers that the problem of "agent memory" is acute enough to justify adopting a new, specialized database,a significant hurdle in any infrastructure layer.
The company's public traction is currently framed in terms of capability, not customer logos. The next twelve months will be critical for moving from technical benchmarks to commercial validation. Key milestones to watch will be the announcement of design partners or early production customers, particularly from verticals like digital health, fintech, or gaming, where behavioral event streams are dense and valuable.
Furthermore, the long-term utility of a specialized "agent database" depends on the sustained growth and architectural patterns of the AI agent ecosystem itself. If agentic workflows consolidate around simpler orchestration layers that rely on existing databases, Keewano's wedge could narrow.
The Standard of Care for AI Memory
Today, the standard approach for giving AI agents access to historical context is a patchwork. Developers might log events to a traditional relational database, a time-series store like InfluxDB, or a data warehouse like Snowflake, then write complex middleware to reconstruct sequences and feed them into an agent's context window. This process is often slow, expensive in compute, and limited by the underlying data model, which fractures context across multiple tables.
The patient population here is every engineering team building AI agents that need to reason over time. The disease state is context loss and reasoning latency, where an agent's performance is gated by its ability to quickly recall and understand the past. Keewano's prescription is a unified store where the event is the primary unit of data, and the timeline is the primary index. It is a bet that the future of machine intelligence requires a new kind of memory, one built not for human reporting, but for autonomous reasoning at scale.
Sources
- [PR Newswire, September 2026] Keewano Raises $12M and Launches KeewanoDB, the First Database Built for Machine Reasoning at Scale | https://www.prnewswire.com/news-releases/keewano-raises-12m-and-launches-keewanodb-the-first-database-built-for-machine-reasoning-at-scale-302878769.html
- [Hetz Ventures, Unknown] Hetz Ventures Portfolio: Keewano | https://www.hetz.vc/portfolio/keewano
- [Citybiz, Unknown] Keewano Raises $12M and Launches KeewanoDB | https://www.citybiz.co/article/903158/keewano-raises-12-million-launches-database-built-for-ai-agent-workloads/
- [Keewano, retrieved 2026] Keewano | The Database for Machine Reasoning at Scale | https://www.keewano.com/
- [Gamigion, Unknown] Article referencing Plarium backgrounds | https://www.gamigion.com/
- [FundUp, Unknown] FundUp profile on Keewano | https://fundup.ai/recently-funded-startups/company/429ad34fdfa2425a141dd8f946831db3177f222fe48fa1a65e4dde29aaa6e356/keewano
- [a16z Speedrun, Unknown] a16z speedrun profile references | https://www.a16z.com/speedrun
- [Hetz Ventures Careers, July 2026] Head of Developer Relations (AI Infrastructure) job posting | https://careers.hetz.vc/companies/keewano/jobs/76297525-head-of-developer-relations-ai-infrastructure