Kirha Monthly Recap — January 2026

We will open our San Francisco office Q1 2026 to meet the rising agentic demand from both Vertical Agent builders and Enterprise clients. The former need a wide supply of curated datasets, the broader your audience, the more data points they’ll query, while the latter need privacy and context auditability. In January we laid solid foundations to prepare for the Agentic Tsunami ahead, while our CEO was learning to surf in Half Moon Bay. Enjoy this Kirha monthly recap.

What We Shipped

Open-Source Tool Planning - Models & SDK

The more we talk with enterprises, the clearer it becomes: privacy is critical. Many are building internal AI tools and strictly forbid employees from using external solutions. Some brave souls still take the risk and use external AIs to get their work done. Why? Because enterprise AI search has been lagging behind tools like Perplexity and Exa. But those large-scale web indexers can’t be deployed on-prem. That’s a real blocker for adoption. If internal AI can’t search the outside world well, usage will stall.

Kirha’s open-source Tool Planning models are now live on Hugging Face to support our on-premise offering. They are compiled, runtime-optimized versions of a fine-tuned Qwen3 8B model that transforms natural language queries into full tool execution plans in a single pass. Read the full docs here. These models are small enough to run on a personal computer. A small price to pay to preserve your independence. Anyone - even those with nothing to hide - can search Kirha context graph with full privacy, run our Tool Planning model and fetch curated context using the SDK. If you can’t find the context you’re looking for please write us an email at developers@kirha.com. We add new curated data sources every week to support enterprise AI adoption. The goal is to grow the graph based on real industry demand and make every data point count.

Because nowhere feels safer than home 🏠

Medical vertical

After JPMorgan Healthcare Week, Kirha launched its Medical Vertical, connecting AI agents to high-value medical sources: clinical trials, OpenFDA drug catalogs, PubMed research papers, preprints, patents, and other reference datasets used in medical research and practice. Pierre built Kirha Life on top of this vertical during the Daytona HackSprint in San Francisco. It’s a fact-based medical hotline with privacy by default:

• Voice-to-text runs locally
• Each call spins up an isolated Daytona sandbox
• The system destroys sandboxes at the end of each conversation
• It returns only grounded medical facts

Kirha Life won an award sponsored by ElevenLabs at the hackathon. It’s already generating commercial interest in the medical space, and we’re exploring early deployment opportunities. Check out the source code to see how easy it is to give your AI agents a better alternative to web search.

AdTech vertical

The AdTech vertical is designed to help agents analyze competitors’ advertising presence across sectors, identify positioning patterns, and track high-level trends in digital campaigns. This vertical serves as an initial foundation for structured advertising intelligence, with a clear focus on market-level insights rather than campaign execution.

Business & Ecosystem

Benchmark Impact on usage

In January, we published the Kirha vs Web Search benchmark. It reached close to one million views on socials and provided the first concrete reference for evaluating Kirha in agent-first workflows using an LLM-as-a-Judge methodology. It also drove significant traffic to kirha.com, with roughly 50 new users registering every day in search of real-time context.

Community Growth

Throughout January, Kirha continued to grow across social platforms, with a clear concentration of activity on X (Twitter). We reached 6,000 builders following Kirha, and the platform has become a regular touchpoint for discussions around benchmarks, agent design, and practical implementation questions.

IRL Events & Network Expansion

January also included several important in-person moments across the US and Europe.

In San Francisco, Pierre met with Dev Chandra, secured a last-minute speaking opportunity to a Pitch competition and won the event.

In Paris, Adam Rothstein and Nicolas Morel attended the Non-Fungible Leaders Conference, engaging with a broad range of actors from the Blockchain, AI, and IT ecosystem. They also hosted a hands-on workshop on AI agents, demonstrating how Kirha improves agent outputs and delivers measurable performance gains in real workflows, with discussions involving private-sector and institutional players, including those focused on ethical and responsible AI.

Closing Thoughts

January marked a transition for Kirha, moving from early visibility and experimentation toward clearer signals of usage and adoption. Benchmarks, community engagement, and initial integrations are converging into real agent workflows, where access to structured, verifiable context matters. Every AI product of the decade will rely on context infrastructure. We’re defining the category with Context-as-a-Service (CaaS).