Dmitry Grankin | Vexa.ai | Meeting API

Founder & CEO at | Open Source Real-Time Meeting Bots API

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

I flipped the switch: Vexa Open Source APIs - real-time Google Meet transcription is live on @Vultr for builders.

Use Hosted to build features, not manage GPUs. Same Apache-2.0 code as self-host.

Subscribe to support open-source work if Vexa’s useful. Comment for $50 credit.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

@AlexSDGuitarist @otter_ai you couldn't kill it because the bot answers to otter, not the host. self-host the capture and the host owns the kill switch — nothing leaves your network. board call is a rough place to find that out.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

@johnjanuszczak @meetgranola "compliant" and "sends your transcript to our cloud" don't really mix. only works self-hosted, where audio + transcript never leave the practice. ours is open source if you want to poke at it.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

The part Whisper doesn't cover is capture: getting clean audio out of Zoom/Meet/Teams in the first place. That's the closed piece people are actually paying Otter/Recall for. We open-sourced that layer too — a bot joins the call → your own Whisper → your own DB. Self-host the whole path, no middleman. Repo: https://t.co/GeL3GjPFCU

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

v0.10.6 shipped last week.

Recording assembly moved out of the bot, into a separate service that reads chunks already in MinIO. Bot can OOM, SIGKILL, get evicted. Recording still finalizes.

Failure class → recoverable event.

Same release deleted ~2,300 lines of platform-duplicated audio code. One module now handles Meet, Teams, Zoom. The dedup is what keeps the guarantee from drifting back into per-platform failure within a quarter.

Apache 2.0. Self-host. Audit trail is the commit history.

https://t.co/ArwEVE2E8B

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

Lever #3 is the cleanest external statement yet of what's happening in this category — it's fragmenting along form-factor (bot vs desktop) × license-shape (closed vs OSS) axes.

Granola's choice (desktop, closed) is one legitimate response. Recall is another (bot, closed). Vexa is a third (bot, Apache 2.0, self-host).

Different buyer-shape for each. The interesting thing is that lever #3 makes the choice visible.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

@pumfleet @calcom Any closed source, presumedly including https://t.co/JydX0aDYQD is still just a glue sticking together a bunch of huge open source projects. And you still depends on the security of the underlying stuff. This will not make https://t.co/JydX0aDYQD secure.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

@LiteLLM @GitHubSecurity still no response - a popular open source project is blocked fo no good reason, github stays silent

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

My GitHub account was compromised in the TeamPCP supply chain attack — attackers used ~73 stolen tokens to spam a security issue on @LiteLLM (BerriAI/litellm#24512).

GitHub flagged my account (the victim) and my open-source org has been returning 404 for 2 days.

Vexa (https://t.co/GeL3GjPFCU) is an Apache 2.0 project with active users (1.8k stars) who are now blocked.

No response on support ticket #4199834. @github @githubsupport @ishaan_jaff @krrish_dh

Can someone please escalate?

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

this is the right framing. CLAUDE.md turns claude from a stranger to a teammate who knows the codebase.

the gap i keep hitting: it still doesn't know what was decided in yesterday's call. you can document every convention, every pattern, every architectural decision — but the "why" behind those decisions usually lives in conversations that never made it into any file.

code context is solved. conversation context is the next frontier.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

the linear issue ingestion is smart — closing the loop from "someone said we should do X" to an actual tracked item. that's the part most teams lose. decisions get made in calls, action items get written in meeting notes nobody reads, and then someone creates the ticket from memory two days later.

meetings → markdown → vault → linked issues across agents is the full pipeline. curious how you handle the meeting → markdown step — manual notes or automated transcription?

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

exactly. and the irony is that "wasn't in any of the meetings" is solvable right now — the tooling exists to pipe transcripts directly into agent context. but most teams treat meeting data as this separate silo that never touches their dev workflow. the chief of staff has access to every file, every PR, every commit... but not the conversation where someone said "actually let's not do it that way."

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

landed on the same pattern - git-backed markdown, agents read/write as they go, not at session end. the part that still bites: what feeds the vault? code diffs, tickets, docs are easy to pipe in. but the decisions made verbally in meetings just vanish. that's the context source nobody's ingesting yet.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

the fact that call transcripts live in the same repo as the slash command is the most underrated part of this whole setup. everyone else optimizes the prompt, you optimized the context pipeline. curious how you get the transcripts in there though - manually copy-pasting from zoom recordings, or do you have something automating that step?

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

@tonybuildsai @Al_Grigor exactly right. the harder version of this problem: the agent doesn't know "don't touch prod until after the demo" because that was said in a call, not written in a file.

code context is solved. conversation context is the gap.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

skills solve the "what does my agent know about this codebase" problem perfectly.

the unsolved one is "what does my agent know about what was decided in yesterday's standup?" conversation context — decisions, customer feedback, verbal agreements — doesn't live in any repo.

that's the next context frontier.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

@0xTib3rius agreed. the meta-question is what type of context has the most impact.

code, docs, and tickets are well-solved. the gap is conversation context — what was decided in standup, what the customer said on the call.

that's the context source agents can't access yet.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

nice choice going with parakeet for diarization. we're evaluating parakeet as a whisper replacement for real-time meeting transcription — the CTC architecture means no hallucinations during silence, which is whisper's biggest production issue.

curious about your experience with speaker embedding quality on short segments. in real-time we're working with 1-3 second chunks and diarization accuracy drops fast at those lengths.

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

this is why we open-sourced our MCP server from day one. Apache 2.0, not "open" until the platform decides otherwise.

we build meeting bot infrastructure — same pattern. competitors ship closed-source MCP servers for their APIs. when the platform changes the rules, their users have zero leverage.

self-host it, fork it, own it. that's the only MCP server nobody can kill.

https://t.co/GeL3GjPFCU

[Danila Poyarkov @dan_note]

Figma shipped a silent patch specifically to kill figma-use — my open-source tool that did what they wouldn't: an MCP server that creates and modifies designs, JSX export, design linting. Then they scrambled to catch up with their own MCP server.

So I spent the weekend recreating @Figma from scratch.

OpenPencil: reads and writes .fig files, AI chat with full design tools, P2P collaboration with zero servers, ~7 MB app. No account, no subscription.

Three days, one developer, MIT license.

https://t.co/bPtP6JPbq0

[Dmitry Grankin | Vexa.ai | Meeting API @grankin_d]

sachin's /save-meeting skill is the one that caught my attention. he downloads meetings from granola into local markdown so claude code can reference what was discussed.

we built vexa to make that pipeline automatic — bot joins the meeting, transcribes with speaker labels, serves it via MCP so any agent can query "what did the team decide about X?"

the 1,500 PMs learning claude code skills will eventually hit the same wall: the agent knows your codebase but not what you decided in yesterday's standup. meeting context is the missing skill category.

[Sachin Rekhi @sachinrekhi]

Yesterday 1,500 product managers joined my live webinar on Claude Code. I went deep into:

- Why I believe Claude Code is the most productive AI platform for PMs
- Showed off the 13 skills I've built to automate workflows across product strategy, design, and execution
- Walked through exactly how to get started by installing Claude Code as well as picking the right set of editors, terminals, and voice tools
- Give away my detailed playbook to build your own skills to automate your own product workflows

If you missed the session, the video is now up on YouTube.

My favorite comment after the session:

"I just finished digesting what Sachin shared, and… wow. Awe beat angst. The thing that excites me most about AI is its potential to free humans up for higher-order work — and how that extra capacity and access could help level the playing field (if we do this responsibly). Huge kudos to Sachin for making that end goal feel so approachable."
- Mathew Kahansky

https://t.co/gxhA8zVncm