David Farah

Building the Web for agents - CTO @

[NATS.io @nats_io]

New Community Guest Post 🔥

"With NATS: the hard parts already exist"

Check out our latest guest blog post from the creators of Cotal: "Coordinating Teams of AI Agents in Real Time on NATS and JetStream".

https://t.co/CgtFe3ItKc

[David Farah @DavidFarahlb]

@imbabybrooklyn We would love to have an official partnership between https://t.co/JubI4FRt77 and Hermes. We are all in on distributed agentic systems and programmable workflows, if this aligns, let me know! DMs open, or [email protected]
@NousResearch

[David Farah @DavidFarahlb]

@RandyHaddad6 Amazing view

[David Farah @DavidFarahlb]

@RandyHaddad6 Very interesting, how do you deal with context window overflow?

[Cotal @Cotal_AI]

Your coding agents can finally work as a team. Across vendors.

Claude Code, Codex, OpenCode and Hermes in one shared space, coordinating so you get the best out of each one.

One command and you're live:
npx cotal-ai setup --full

What's your multi-agent setup? Show us.

#ClaudeCode #Codex #OpenCode #Hermes #AIagents

[David Farah @DavidFarahlb]

@0xRicker I think this would be useful to you: https://t.co/8SS3hLQZd5

[David Farah @DavidFarahlb]

@Cotal_AI Excited for this new protocol!
Long running multi agent systems are on the horizon and Cotal makes it easy to create and try out different topologies

[Cotal @Cotal_AI]

The web for agents is here!

COTAL is the open protocol that lets your agents coordinate with each other.

One protocol to connect them all!

Try it in one command:
npx cotal-ai setup --full

open standard, Apache-2.0
https://t.co/5M1750j7t8

[David Farah @DavidFarahlb]

@ishaansehgal This is very interesting, thanks for sharing.
This might help you out with coordination, check it out: https://t.co/8SS3hLQZd5

[David Farah @DavidFarahlb]

@kaygelly Would love to hear more about it

[David Farah @DavidFarahlb]

@Walterwu44739 @frontiertower @svensonj00 Thanks for the interview Wu! check out our repo here: https://t.co/8SS3hLQZd5

[Wu | HiRey.ai @Walterwu44739]

What are you building? #32

met @svensonj00 and @DavidFarahlb at @frontiertower.

They are building Cotal: a communication layer for agents.

Their line: protocol for WebA, one protocol to connect them all.

The idea is that WebA cannot just be one big orchestrator controlling a bunch of sub-agents.

Long-running agents should be able to talk to each other, pass messages, coordinate work, and avoid stepping on each other.

Sven and David came from Switzerland to SF because they heard people here are building WebA.

That line felt very SF to me: you hear a new internet layer might be forming somewhere, so you get on a plane.

Let's build WebA together.

[David Farah @DavidFarahlb]

@svensonj00 Great!

[David Farah @DavidFarahlb]

@Wanawrin Great job!

[Filip M @Wanawrin]

I made my agent 34.2% more accurate by letting it self-improve. 📈

One command, thats it. Full explanation below.

[Tony @halluton]

just used voice mode to self-improve my agent in 20 seconds

talked to claude code with /voice, it pulled insights from past agent failures via @KaybaAI & directly implemented the fixes

[David Farah @DavidFarahlb]

@halluton @KaybaAI Impressive!

[Kayba @KaybaAI]

Kayba live on @ProductHunt

- agents fail the same way every run and never learn
- kayba analyzes traces and extracts what went wrong
- your coding agent deploys the fixes automatically

https://t.co/DXaBH5kOJm

[Kayba @KaybaAI]

benchmarked on τ2-bench: up to 2x improvement in agent consistency

the dashboard is built to make your agent self-improving out of the box. DM for a free 7-day Pro trial

the underlying framework stays open source and free forever: https://t.co/ZTspVAU66V

[Kayba @KaybaAI]

we turn traces into self-improving agents:

1/ analyze: upload traces, kayba finds what works and what doesn't
2/ review: every skill links back to its source conversation. you decide what stays
3/ deploy: approved skills become a prompt appendix. paste into your system prompt