Connery

Open-source plugin infrastructure for AI.

⭐ GitHub →

[Connery @connery_io]

Interesting, what happens when you let two AIs correct each other. Boosts precision. Pretty important when you chat with your database in Slack.

Connery's newest use case. Help us build this and join for early access, all for free.

https://t.co/sOm6fqnLDJ

[Michael Liebmann @LiebmannMichael]

Two AI models working together in Slack to chat with your database
- Connery AI assistant: plans and interprets results like a co-pilot and calls AI actions
- SQL generating AI: gets DB schema, creates SQL and runs it
- Self-correction mechanism
- Data drill-downs with chat memory
- Open and extensible
- 100% @SlackHQ, no new tools

More details: https://t.co/JMnLCQmZDv

Signup to our beta waitlist for early access. Love to hear your thoughts and feedback.

[Connery @connery_io]

The technical issues can be overcome. More important are the benefits to business users, having a co-pilot to walk through the analysis and plan and reason together, not just write queries and execute them.

#AI #SQL #DataScientist

[Michael Liebmann @LiebmannMichael]

“Don’t trust AI to write SQL,” they say.

Because results are unreliable? Because business owners don’t know how to request data? That’s NOT true.

We got it working accurately - inside @SlackHQ . No extra tools. Just chat & query.

Watch how @connery_io calculates subtotals from natural language. ⬇️

💡 Join the beta & get early access → https://t.co/lOr6Mtdktd

[Connery @connery_io]

How to increase precision having an AI write and run SQL reading actions:

[Michael Liebmann @LiebmannMichael]

One AI Writes SQL. The Other Fixes It. No human intervention.

AI-generated SQL can fail for many reasons. In @connery_io we found a way to increase this precision bump.

It reads the error, fixes the query, and re-runs —automatically.
No human needed. Just results.

How it works:
1️⃣ Ask Connery for data from your DB in @SlackHQ .
2️⃣ Connery calls SQL-gen AI that writes & runs query.
3️⃣ Query fails. (Syntax issue? Wrong column?)
4️⃣ Connery self-corrects inputs & re-runs ✅

There us much to unpack here. Will add more context later. We let this self-correction only run on READ actions. WRITE is a whole different game.

🚀 Beta is live! Try it now
https://t.co/lOr6Mtdktd

#AI #SQL #Slack #Automation

[Connery @connery_io]

The key here is scoping+HitL:

1. Limiting the scope of changes an AI-generated WRITE query can make.

2. Human-in-the-Loop to approve those (limited) changes, before execution.

All that from Slack with non-tech users.

[Michael Liebmann @LiebmannMichael]

Who dares to have an AI make DB changes based on a Slack chat?

We built the AI assistant '@connery_io’ for Slack that lets you chat with your database – and it also WRITES changes into it.

You can now define AI-powered database actions from @SlackHQ that allow controlled READ/WRITE access for – so AI can update data, but only within the boundaries you set.

Why This Matters
🚯 No more manual SQL edits for routine updates.
♻️ Business teams can self-serve without risking unintended DB changes.
👀 Automation with human-in-the-loop — keep humans in control while speeding up workflows.

How It Works
1️⃣ Query: Ask AI assistant a question in Slack (“Show me all customers on a free plan.”)

2️⃣ Action: Request AI to make a change to your DB, like:
- Update customer addresses based on CRM records
- Change subscription plans for a segment of users
- Mark unpaid invoices as reviewed after approval

3️⃣ Human-in-the-Loop Approval: Approve AI-detected changes to identified record(s).

Key Features
✅ Scoped Permissions – You define what data AI can read or edit.

🔄 Self-Correcting SQL – AI adjusts queries if they fail.

🛠️ Customizable Actions – Create safe, repeatable workflows inside Slack.

📊 Drill-Downs & Reports – Combine read + write actions for deeper insights.

We’re looking for beta testers to explore AI-assisted DB updates in Slack. If this sounds useful, join the beta here:
👉 https://t.co/JMnLCQmZDv

What database actions would save your team the most time? 🚀

[Connery @connery_io]

Next Connery plugin coming up for #mySQL

[Lien @LienLiebmann]

@LiebmannMichael This is great (and for postgress) but I need mySQL. Add for me pleeeeease 🙏🙋‍♀️

[Connery @connery_io]

@ollama SDK for AI actions https://t.co/Jiu3FJK9Pe

[knishimae @knishimae0531]

Cool tutorial from the Connery team on how to use their platform to add arbitrary actions to OpenGPTs!
https://t.co/31l27ZWs81
https://t.co/v0crqsnNfo

[LangChain @LangChain]

🔌 Connery Plugins + OpenGPTs 🦜🤖

👇 Blog + video links at the end!

Some of the coolest LLM-powered apps involve granting the model agentic access to third-party systems, but there are a few things to think about when shipping:

👥 Personalization: How do you manage linked integrations for your users?
🔒 Security: How do you ensure your users' credentials are secure?
🦺 Safety: How do you ensure the LLM doesn't abuse the power you give it?

@connery_io is an open source plugin infrastructure that aims to address these considerations and make it easy to write and share plugins. @LiebmannMichael wrote up a guest blog post and made a video walkthrough integrating them with OpenGPTs!

Blog: https://t.co/Cv7ATjPXRv
Video walkthrough w/ OpenGPTs: https://t.co/qEZdnbb37O

[Connery @connery_io]

We're thrilled to announce that our Slack app is now open source! 🌟

Check out the repository here: https://t.co/zyXwe1WPUf

[Connery @connery_io]

This Connery plugin allows sending SMS and Email notifications from your AI agent using @novuhq

[Connery @connery_io]

We've got a new community plugin: "Novu Send Email/SMS Plugin" (https://t.co/RuSpyil9z9).

Thank you, intertwine, for the contribution!

[Connery @connery_io]

Many thanks, @Hacubu, for mentioning, support and a really fast release! 🙏

[Connery @connery_io]

Connery Toolkit is available on LangChain JS!

Now, you can use Connery actions in your agents. And Connery will handle the rest: isolated runtime, authorization, secret management, and many more.

Integration with LangChain Python is coming.

[Jacob Lee @Hacubu]

🤖 Connery Actions Tool for Agents 🤖

@connery_io is an OSS plugin infrastructure you can use with @langchain JS/TS agents in 0.0.203!

Define a set of actions using integrations like Gmail, and let them handle auth + secret management.

s/o machulav!

https://t.co/LOD8TiE03K

[Michael Liebmann @LiebmannMichael]

We spent months building an open source API wrapper called @connery_io to create/run actions for AI assistants. Adding auth, security, and access management.

Last week @OpenAI made it super easy to add Connery to GPTs. See a Gmail example in the video

Let me know what you think

[Connery @connery_io]

7/ ✨ We welcome your thoughts and are eager to assist. Your input drives our platform forward! Stay connected for more updates.

[Connery @connery_io]

🚀 Excited to introduce Connery Platform - an Open-source plugin ecosystem for AI and No-code.

Now available on GitHub with an MIT license: https://t.co/BX6L2pdVM1

#opensource #ai #NoCode

[Connery @connery_io]

7/ 🚀 Ready to jumpstart your journey with the Connery Platform? Get moving with our QuickStart guide: https://t.co/Jsq3VHYVvN