Israeli startup Keewano has closed $12 million in total seed funding to ship a database that stores events in the order AI agents need them.

The Tel Aviv company is pitching KeewanoDB at what it calls the gap between human-facing dashboards and machine-facing reasoning. Hetz Ventures led the round, with a16z speedrun, Remagine Ventures, DIG Ventures and unnamed angels participating.

Co-founder and CEO Mark Kardashov set the architectural ambition plainly. "In 1964, Japan didn't try to make trains faster on the same track. They built a new line," he said. "Databases are at that fork now. AI agents need a fundamentally different data architecture, and we've built KeewanoDB as the new line."

What problem does the database solve?

Keewano argues that conventional analytics warehouses were designed around a human typing SQL into a box. Events land as flat rows, columns, and precomputed aggregates; reconstructing the sequence in which things happened requires reassembly at query time. The cost scales with event variety, so most production systems capture only a few hundred distinct event types before performance erodes.

Kardashov described the founding motivation in operational terms. "We were trying to use AI agents to answer business questions like 'Which of our users are on the same path as the ones who churned last month?' You could see what happened on any dashboard, but why it happened lived somewhere the databases couldn't reach," he said. "Our databases had been designed for humans querying tables, not machines that need to reason live, with complete context."

How does KeewanoDB differ from a conventional warehouse?

The core design keeps each entity's full event sequence stored contiguously and in chronological order on disk. An agent can read a causal chain directly from the raw bytes rather than rebuild it across joins. Because the layout is alignment-friendly, the engine can process roughly ten events per CPU instruction using SIMD instructions, according to the company.

A semantic layer maps raw event names and field labels to shared concepts. Clients connect through MCP, SQL, REST, and conventional APIs. Keewano claims a single query can scan a quarter of a billion events in half a second and return context already structured for downstream reasoning.

What does the commercial model look like?

Keewano prices on monthly active users rather than per-event ingestion or compute. The company argues this decouples capture cost from analytical ambition; teams can retain more telemetry without pricing penalties for scale.

The free tier offers up to 3 million events and a ceiling of 27,000 monthly active users per cycle. Projects that exceed either limit go read-only. Stored data remains queryable, but new ingestion pauses until the customer upgrades. A fully managed Keewano Cloud service can sit alongside an existing warehouse or replace it through standard integrations.

Who is behind the company?

Keewano was founded in 2024 in Tel Aviv by four people:

• CEO Mark Kardashov
• COO Dima Karger
• CTO Pavel Bilbergal
• Chief Product Officer Vitaly Bukhovsky

Bukhovsky and Kardashov previously co-founded TestProject, a cloud test automation platform acquired in 2015, and Devalore, a software firm sold in 2022. Karger and Bilbergal previously worked together at Plarium on game engines, performance, and multi-platform systems serving millions of players.

How does this fit the broader AI infrastructure market?

The pitch intersects with a wider push to retool the data stack for inference workloads. Established warehouses remain optimised for batch analytics and dashboarding, and vendors from Snowflake to ClickHouse are racing to add agent-friendly interfaces such as MCP support. Keewano's wager is that incremental retrofitting will not satisfy agents that need ordered, full-fidelity event context on every query.

If the order book grows beyond early design partners, the SIMD-first approach and MAU pricing will come under direct scrutiny from incumbents racing to ship similar primitives. The next test for Keewano is whether paying customers will swap, or augment, established warehouses once their AI agents start asking the harder questions.