Your meetings are a goldmine of knowledge. Start mining.
Every meeting tool gives you individual meeting notes. Only Notemesh gives you collective intelligence — the ability to ask questions across all your meetings and get cited answers in seconds. This is the feature that changes how teams think about meetings.
The problem with meeting notes
Your company holds hundreds of meetings every month. Each one contains valuable information: strategic decisions, client feedback, technical explanations, competitive intelligence, process knowledge, and institutional memory. But all of it is trapped in individual meeting notes that nobody ever reads again.
When someone new joins the team, they spend weeks asking questions that were answered in meetings before they arrived. When a client references something discussed six months ago, your team scrambles to reconstruct the context. When leadership asks "what have customers been saying about feature X?", the answer requires manually reviewing dozens of call recordings.
The knowledge is there. It is just inaccessible. Until now.
How it works
Tag meetings into knowledge bases
Create tags for clients, projects, topics, departments, or any organizational structure that makes sense for your team. Assign meetings to tags manually, or let Notemesh auto-suggest tags based on meeting content. A single meeting can belong to multiple tags — a client call might be tagged under both the client name and the project name.
Automatic embedding and indexing
When a meeting is tagged, Notemesh automatically splits the transcript into semantic chunks (200-500 words by speaker turn), generates vector embeddings using OpenAI's text-embedding-3-small model, and stores them in a pgvector database. This happens in the background — you assign the tag and Notemesh handles the rest.
Ask questions with AI chat
Open any knowledge base and start chatting. Type a question like "What pricing concerns has Acme raised over the last quarter?" and the AI performs a vector similarity search across all meeting chunks in that tag, assembles the most relevant context, generates a coherent answer, and includes citations linking to exact timestamps in the source meetings.
Verify with citations
Every answer includes citations in the format [Meeting Title, MM:SS]. Click any citation to jump directly to that moment in the original meeting recording. This is not a black box — every claim the AI makes can be verified against the source material. Trust the AI, but verify when it matters.
Key benefits
Ask questions across hundreds of meetings
"What were the top three feature requests from enterprise clients this quarter?" "When did we first discuss moving to microservices?" "What did the legal team say about the data retention policy?" Questions that would take hours to answer manually are answered in seconds with cited sources. The more meetings you record, the smarter your knowledge base becomes.
Answers with verifiable citations
Every answer includes clickable citations that link to the exact meeting and timestamp where the information was discussed. This is not summarization of summaries — it is RAG-powered retrieval from original transcripts. When you need to verify a claim, share evidence with a stakeholder, or trace a decision back to its origin, the citation takes you directly to the source.
Project plans with milestone tracking
Knowledge bases can generate project plans by synthesizing timelines, milestones, and commitments across all tagged meetings. Ask "Create a project timeline based on all meetings tagged under Project Phoenix" and get a structured plan with dates, owners, and dependencies — all sourced from actual meeting discussions, not someone's best guess from memory.
Shared knowledge for teams
On Team plans, knowledge bases are shared across invited team members. A sales team can build a knowledge base per client that every account executive can query. An engineering team can build one per project that new developers can search during onboarding. Knowledge stops being locked in individual heads and becomes a team resource.
What makes this different from every other meeting tool
Other meeting assistants give you a transcript and a summary for each individual meeting. That is useful, but it is only the first step. The real value of meeting data is not in any single meeting — it is in the patterns, decisions, and knowledge that emerge across dozens and hundreds of meetings over time.
Notemesh's knowledge base is the only feature on the market that lets you treat your meeting history as a queryable database of institutional knowledge. It is the difference between having a filing cabinet full of notes and having a research assistant who has read every note and can answer any question about them.
This is not incremental improvement. This is a fundamentally different way to think about meetings. Every meeting you record makes the knowledge base smarter, more comprehensive, and more valuable. The longer you use Notemesh, the wider the gap becomes between your team's institutional knowledge and everyone else's.
Use cases
Institutional knowledge retention
When a senior engineer leaves, years of context walk out the door. With Notemesh, every technical discussion, architecture decision, and design rationale from their meetings is preserved in the knowledge base. New team members can query it to understand why things were built the way they were — even years later.
New employee onboarding
Instead of spending three weeks in "context download" meetings, new hires can query the knowledge base. "What are the current product priorities?" "Who are our top five clients and what are their main pain points?" "What was decided about the Q4 roadmap?" Answers come from actual meeting discussions, not stale wiki pages.
Deal intelligence for sales
Tag all meetings with a prospect under their company name. Before the next call, ask the knowledge base "What concerns has this prospect raised?" "What competitors have they mentioned?" "What features are most important to them?" Walk into every call with perfect recall of everything discussed in every previous meeting.
Research synthesis
UX researchers, product managers, and analysts conduct dozens of interviews and feedback sessions. Tag them all into a research knowledge base and ask "What are the most common usability complaints?" or "How do enterprise users differ from SMB users in their workflow?" Get synthesized insights with citations to specific interview moments.
Under the hood
Vector embeddings
Transcripts are chunked into 200-500 word segments by speaker turns and embedded using OpenAI's text-embedding-3-small model (1536 dimensions). Embeddings are stored in PostgreSQL with pgvector for fast cosine similarity search.
RAG architecture
Questions are embedded with the same model, matched against stored chunks via vector similarity, and the top results are assembled into context for the AI. The AI generates answers constrained to the provided context with explicit citation instructions.
Batch processing
Embedding generation is batched (20 chunks per API call) and processed asynchronously via BullMQ job queues. Tagging a meeting triggers automatic embedding — no manual step required.
Citation extraction
Citations are parsed from AI responses using the [Meeting Title, MM:SS] pattern and linked to the source meeting and timestamp. Click any citation to jump to that moment in the recording.
Frequently asked questions
What is a meeting knowledge base?
A meeting knowledge base is a tagged collection of meetings that Notemesh indexes and makes searchable via AI chat. You create a tag — like "Acme Client," "Product Roadmap," or "Engineering Standup" — and assign meetings to it. Notemesh then embeds every transcript chunk into a vector database, allowing you to ask questions that are answered by drawing from all meetings in that collection. Think of it as having a research assistant who has attended every single meeting and can recall anything instantly.
How does the AI chat with citations work?
When you ask a question in a knowledge base, Notemesh embeds your query, performs a vector similarity search across all meeting transcript chunks in that tag, assembles the most relevant context, and sends it to the AI along with your question. The AI generates an answer and includes citations in the format [Meeting Title, MM:SS] so you can verify every claim by clicking through to the exact moment in the original recording.
How many meetings can a knowledge base contain?
There is no practical limit on the Pro or Team plans. Knowledge bases can contain hundreds or even thousands of meetings. The vector search is designed to scale — a query against a knowledge base with 500 meetings returns results just as fast as one with 5 meetings, because the search operates on pre-computed embeddings rather than scanning raw text.
Can I share a knowledge base with my team?
Yes. On Team plans, knowledge bases (tags) can be shared with team members. When you share a tag, all meetings assigned to it become accessible to invited members. You control access through team roles and departments. Shared knowledge bases are particularly powerful for sales teams, project teams, and cross-functional groups that need collective access to meeting intelligence.
What makes this different from just searching transcripts?
Full-text search finds exact or similar words. Knowledge base AI chat understands concepts. Ask "What were the main objections the client raised about pricing?" and the AI synthesizes answers from across dozens of meetings where pricing was discussed — even if the word "objection" was never used. It understands semantic meaning, connects related discussions across time, and generates coherent answers with source citations. This is the difference between a search engine and a research assistant.
Related features
Related articles
Your competitors are losing meeting knowledge. You do not have to.
Notemesh is the only meeting assistant that turns your meeting history into a queryable knowledge base. The longer you wait, the more institutional knowledge you lose. Start today.
Try Notemesh Free →