Composio
974 posts
Composio
@composio
Your agent is smart. Its tools should be too. Check https://t.co/KuJTCSKgDi
build on top of us ➡️ Katılım Ekim 2023
51 Takip Edilen16.1K Takipçiler
Your AI agent is only as good as the tools it can actually reach.
"If you're building a deep research agent that needs to pull from Notion, we want to make sure all your integrations are sorted first. We loop in the Composio team directly. Shared channels between the customer, us, and Composio."
@contextkingceo (founder of @hydra_db) on why the integration layer has to come before the retrieval layer and why @Composio is the first call they make.
Full conversation below + on YT and Spotify!
Julia Fedorin@juliafedorin
Vector databases are a scam. Not technically, they do exactly what they say. Return the most cosine-similar string to your query. The scam is the entire industry pretending that's the same thing as relevance. It isn't. Search "Apple." You get the fruit, the company, the watch, and a recipe blog. Your agent picks one at random and calls it retrieval. Your customer calls it broken. Most AI agents shipping right now are duct-taped on top of this. They demo well because demos are easy. They die in production because production is real. @Hydra_db's Founder Nish (@contextkingceo) said the quiet part out loud — "vector databases suck, similarity is not relevance" — and the demo signups haven't stopped since. He raised $6.5M because he was the first to name what everyone in the room already knew. If your retrieval layer is a flat embedding index, you're not building infrastructure. You're building a liability with a prettier name. 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 (00:00) AI Needs Context (01:30) HydraDB Explained (07:41) Vector Search Breaks (09:32) Messaging That Converts (13:41) Writing the Viral Tweet (16:07) Similarity Not Relevance (20:46) POC to Production Gap (35:35) Raising 6.5 Million Fast (39:33) Founder Lesson on Messaging This is a @Composio "Agents at Work" podcast, where I chat with founders building the next leap of AI. Follow for more:)
English
Composio retweetledi
At @composio everyone is vibe-coding and building agents. Growth team included.
So I built an internal agent that roasts every PR in the format of leaked group chats 🤣
PS: If this gets ratio'd I'll ship it.
English
Composio retweetledi
A user told @contextkingceo his website was extremely shitty.
It was the best feedback he ever got.
"Someone signed up because a person told them explicitly what we do. They said — your website is extremely shitty, you don't communicate what you do. No one has enough time to figure it out on your behalf."
So they threw out every flowery, smart-sounding word on their website and replaced it with just:
"Vector databases suck."
Bonus points to Nish for saying something no one else had the guts to say.
200 companies signed up for demos after that.
Full conversation below.
Julia Fedorin@juliafedorin
Vector databases are a scam. Not technically, they do exactly what they say. Return the most cosine-similar string to your query. The scam is the entire industry pretending that's the same thing as relevance. It isn't. Search "Apple." You get the fruit, the company, the watch, and a recipe blog. Your agent picks one at random and calls it retrieval. Your customer calls it broken. Most AI agents shipping right now are duct-taped on top of this. They demo well because demos are easy. They die in production because production is real. @Hydra_db's Founder Nish (@contextkingceo) said the quiet part out loud — "vector databases suck, similarity is not relevance" — and the demo signups haven't stopped since. He raised $6.5M because he was the first to name what everyone in the room already knew. If your retrieval layer is a flat embedding index, you're not building infrastructure. You're building a liability with a prettier name. 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 (00:00) AI Needs Context (01:30) HydraDB Explained (07:41) Vector Search Breaks (09:32) Messaging That Converts (13:41) Writing the Viral Tweet (16:07) Similarity Not Relevance (20:46) POC to Production Gap (35:35) Raising 6.5 Million Fast (39:33) Founder Lesson on Messaging This is a @Composio "Agents at Work" podcast, where I chat with founders building the next leap of AI. Follow for more:)
English
Composio retweetledi
"Similarity is not relevance."
That's the line that explains why your AI agent nails the demo and falls apart in production.
Vector databases embed everything and cosine-similarity it. The problem? Cosine doesn't know what your business actually means.
Search "Apple." Do you get the fruit, the company, or a watch review?
Nishkarsh Srivastava (@contextkingceo), founder of @Hydra_db built an ontology-first context layer that preserves relationships instead of flattening them.
The result: agents that return relevance, not just similarity.
Full episode below and also on YT and Spotify:)
Julia Fedorin@juliafedorin
Vector databases are a scam. Not technically, they do exactly what they say. Return the most cosine-similar string to your query. The scam is the entire industry pretending that's the same thing as relevance. It isn't. Search "Apple." You get the fruit, the company, the watch, and a recipe blog. Your agent picks one at random and calls it retrieval. Your customer calls it broken. Most AI agents shipping right now are duct-taped on top of this. They demo well because demos are easy. They die in production because production is real. @Hydra_db's Founder Nish (@contextkingceo) said the quiet part out loud — "vector databases suck, similarity is not relevance" — and the demo signups haven't stopped since. He raised $6.5M because he was the first to name what everyone in the room already knew. If your retrieval layer is a flat embedding index, you're not building infrastructure. You're building a liability with a prettier name. 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 (00:00) AI Needs Context (01:30) HydraDB Explained (07:41) Vector Search Breaks (09:32) Messaging That Converts (13:41) Writing the Viral Tweet (16:07) Similarity Not Relevance (20:46) POC to Production Gap (35:35) Raising 6.5 Million Fast (39:33) Founder Lesson on Messaging This is a @Composio "Agents at Work" podcast, where I chat with founders building the next leap of AI. Follow for more:)
English
Composio retweetledi
Despite being told no, I'm open-sourcing TrustClaw.
You can now deploy a production-ready personal agent service with over 1000+ app integrations in a single command, straight to @vercel with npx @composio/trustclaw deploy
I was inspired by @openclaw to build a simple web app where anyone could create their own 24/7 personal assistant and connect it to Gmail, Google Calendar, Notion, Slack, GitHub, HubSpot, Linear… well everything, and securely through OAuth/sandbox execution.
It went viral on X, reached over a thousand users in less than 48h, and revenue began pouring in.
If you are thinking like a company, you'd probably keep that locked up. But why should I be the reason you spend another year scrolling instead of building?
So today, I'm open-sourcing TrustClaw anyway.
> 24/7 agents that act across Gmail, Notion, GitHub, Slack, Linear, Jira, and 1000+ apps
> OAuth and sandboxed execution, so users don't have to hand agents passwords or raw API keys
> Supports multiple users and authentication right outside of the box with @better_auth
Repo is open, MIT licensed.
If I were starting an AI company today, I'd clone this, pick a market, and begin shipping with Claude Code.
Honestly so excited to see what comes out of this.
English
Composio retweetledi
Our internship applications just opened 5 minutes ago…
This is an INSANE opportunity to work with @composio at the fronteir of AI agents in San Francisco.
If you want to:
> directly impact 1,000,000s of people with your work.
> get 3 years of experience in 3 months at a top series A startup, working with insanely smart people.
> work in beautiful sf, meals payed for, team outings, well above market pay and visa sponsorship.
Apply using the link in the comments.
Oh, and there’s only one posting because you will get to build your own internship depending on what you want to work on.
English
Composio retweetledi
Must watch
Julia Fedorin@juliafedorin
Vector databases are a scam. Not technically, they do exactly what they say. Return the most cosine-similar string to your query. The scam is the entire industry pretending that's the same thing as relevance. It isn't. Search "Apple." You get the fruit, the company, the watch, and a recipe blog. Your agent picks one at random and calls it retrieval. Your customer calls it broken. Most AI agents shipping right now are duct-taped on top of this. They demo well because demos are easy. They die in production because production is real. @Hydra_db's Founder Nish (@contextkingceo) said the quiet part out loud — "vector databases suck, similarity is not relevance" — and the demo signups haven't stopped since. He raised $6.5M because he was the first to name what everyone in the room already knew. If your retrieval layer is a flat embedding index, you're not building infrastructure. You're building a liability with a prettier name. 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 (00:00) AI Needs Context (01:30) HydraDB Explained (07:41) Vector Search Breaks (09:32) Messaging That Converts (13:41) Writing the Viral Tweet (16:07) Similarity Not Relevance (20:46) POC to Production Gap (35:35) Raising 6.5 Million Fast (39:33) Founder Lesson on Messaging This is a @Composio "Agents at Work" podcast, where I chat with founders building the next leap of AI. Follow for more:)
English
Composio retweetledi
Mobbin MCP is now live on Composio.
rebekah bek@paperseasons
just launched: @mobbin mcp 💛 your AI agents can now search 600,000+ real app screens. paywalls, onboarding, checkout, permissions — from apps that already shipped. ai tools can write code. they just don't know what good looks like. now they do. 👉 mobbin.com/mcp
English
Composio retweetledi
Vector databases are a scam.
Not technically, they do exactly what they say. Return the most cosine-similar string to your query. The scam is the entire industry pretending that's the same thing as relevance.
It isn't.
Search "Apple." You get the fruit, the company, the watch, and a recipe blog. Your agent picks one at random and calls it retrieval. Your customer calls it broken.
Most AI agents shipping right now are duct-taped on top of this. They demo well because demos are easy. They die in production because production is real.
@Hydra_db's Founder Nish (@contextkingceo) said the quiet part out loud — "vector databases suck, similarity is not relevance" — and the demo signups haven't stopped since.
He raised $6.5M because he was the first to name what everyone in the room already knew.
If your retrieval layer is a flat embedding index, you're not building infrastructure. You're building a liability with a prettier name.
𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
(00:00) AI Needs Context
(01:30) HydraDB Explained
(07:41) Vector Search Breaks
(09:32) Messaging That Converts
(13:41) Writing the Viral Tweet
(16:07) Similarity Not Relevance
(20:46) POC to Production Gap
(35:35) Raising 6.5 Million Fast
(39:33) Founder Lesson on Messaging
This is a @Composio "Agents at Work" podcast, where I chat with founders building the next leap of AI.
Follow for more:)
English
.@Morphllm has a hiring framework most companies don't.
Before bringing anyone on they ask one question: is this person worth more than just buying more GPUs?
"We spend 8x more on GPUs than we do on salary."
@Tejasybhakta(Founder of @Morphllm) isn't building a big team. He's building a fast model.
Full conversation on YouTube and Spotify, link in comments
English
The people asked, and @sarahfim delivered. Trustclaw is now open source and can be deployed in a single command.
sarah@sarahfim
Despite being told no, I'm open-sourcing TrustClaw. You can now deploy a production-ready personal agent service with over 1000+ app integrations in a single command, straight to @vercel with npx @composio/trustclaw deploy I was inspired by @openclaw to build a simple web app where anyone could create their own 24/7 personal assistant and connect it to Gmail, Google Calendar, Notion, Slack, GitHub, HubSpot, Linear… well everything, and securely through OAuth/sandbox execution. It went viral on X, reached over a thousand users in less than 48h, and revenue began pouring in. If you are thinking like a company, you'd probably keep that locked up. But why should I be the reason you spend another year scrolling instead of building? So today, I'm open-sourcing TrustClaw anyway. > 24/7 agents that act across Gmail, Notion, GitHub, Slack, Linear, Jira, and 1000+ apps > OAuth and sandboxed execution, so users don't have to hand agents passwords or raw API keys > Supports multiple users and authentication right outside of the box with @better_auth Repo is open, MIT licensed. If I were starting an AI company today, I'd clone this, pick a market, and begin shipping with Claude Code. Honestly so excited to see what comes out of this.
English
Composio retweetledi
Our version of River is Zen. One of our engineers and I built it and it literally accelerated how team works internally.
It runs Claude Code in cloud sandboxes from our Slack, with every tool(Github, Linear, Sentry, GDrive, Gmail, GCal, Notion, our internal APIs) we plugged in through Composio (gotta dogfood our own product). It reads code, runs tests, opens PRs, queries production, pulls customer context. You hand off the task and the PR lands in Github and does more things for us now.
We tried Devin but not having context on our codebase, DB and a bunch of tools we use only added friction.
10 weeks in, Zen completed ~1500 tasks at 94% success rate. averaging 22/day, peaked at 56. 17% of our work happens on weekends now and in the last 2 weeks we saw the highest volume. The thing I didn't see coming. Zen edits its own prompts after each task.
Tobi nails the public-channel-only part. Composio is one example but I predict there will be more startups building Rivers and Zens in the future.
tobi lutke@tobi
English
Powered by Composio!
TinyHumans AI@tinyhumansai
Stoked to announce @composio is now the tool-calling layer inside us 👀 > OAuth handled end-to-end > 1,000+ apps available on demand. you just name it > tools resolved by intent, not config > every call writes back into the memory engine check it out: github.com/tinyhumansai/o…
English