Almanac: The AI Assistant That Knows Your Company Inside Out
YC-backed startup launches context-aware AI agent that automates tasks based on your company's knowledge base.
1The Problem with Current AI Assistants
Most AI tools treat memory as an afterthought, requiring users to repeatedly provide context. The Almanac founders discovered this pain point firsthand while building internal tools for their own company. They saw YC batchmates struggling with the same issue - agents that couldn't maintain context across sessions or understand organizational knowledge. This realization led them to focus on building an AI with proper memory architecture from the ground up.
2How Almanac Works
Almanac creates two dynamic wikis - a personal one tracking individual preferences and relationships, and a company wiki containing organizational knowledge like roadmaps and blockers. The AI agent accesses these wikis plus connected accounts (Gmail, Calendar, etc.) through secure OAuth connections. Unlike session-bound assistants, Almanac maintains continuous context, enabling it to pick up tasks days later without losing the thread. The system runs background workers that proactively identify automatable tasks.
3Proactive Automation
Almanac's pre-compiled knowledge base enables truly proactive assistance. The system can draft fundraising pitch decks before being asked, follow up on emails hours after they're sent, and maintain context on long-running projects. One user runs their dog-rescue operation through Almanac, automating foster coordination and consent form reminders. Another uses it for Polymarket strategy research, comparing past performance to suggest new approaches.
4Technical Differentiation
The team's key innovation is their approach to memory architecture. While most AI tools add memory as an afterthought, Almanac was built around pre-compiled knowledge layers from the start. This stems from the founders' Harvard capstone thesis on pre-compilation. Their system invests significant upfront compute to properly structure organizational knowledge, enabling more reliable context-aware operations than typical chatbot-style assistants.
5Founders' Background
The three co-founders - Kushagra, Rohan, and Divit - have been friends for 11 years since preparing for India's competitive IIT-JEE exams. All studied Electrical Engineering (at IIT Delhi, IIT Kharagpur, and BITS Pilani) before Kushagra and Rohan attended Harvard. Their previous work includes building knowledge layer products for Harvard and NASA. This technical pedigree informs their unique approach to AI memory systems.
6Privacy Architecture
Almanac only accesses explicitly connected accounts through OAuth providers, with credentials never stored in its database. The system retains cited information as Markdown rather than original documents. Users can create personal accounts (visible only to them) and shared company accounts. The architecture ensures founders can't access each other's private accounts while maintaining collaborative company wikis.
7Diverse Use Cases
Early adopters are using Almanac in unexpected ways beyond corporate productivity. One marketing professional builds campaigns without re-explaining business context each time. A trader researches investment strategies by comparing historical performance. The system's ability to maintain specialized knowledge makes it adaptable across domains - from dog rescues to financial markets to creative workflows.
8The Future of Work AI
Almanac represents a significant evolution from today's session-bound AI tools. By properly architecting organizational memory and enabling continuous context, it moves closer to how humans actually work - across days, with interruptions, and multiple concurrent projects. As companies adopt such systems, we may see fundamental shifts in how knowledge work gets done, with AI handling the cognitive overhead of context maintenance.