Ship agents. Not auth infrastructure. | github.com/ArcadeAI
San Francisco, CA
- Tweets 293
- Following 30
- Followers 1,011
- Likes 452
Open-source Slack agent that connects to @github, Gmail, and Calendar.
Built with @modal_labs + @langchain
Check out the repo: github.com/ArcadeAI/SlackAge…
8
41
9,448
🚀 Just launched: Arcade Deploy
Build custom AI tools that connect to your database, Salesforce, or APIs—then deploy with a single command.
No servers to manage. No infrastructure to configure. No deployment pipelines.
From code to production in seconds.
Watch how ⬇
2
2
20
1,300
@langchain + @TryArcade = Better Agents
Now you can use Arcade tools within Langchain/graph or authorize existing Langchain tools
LangGraph integration works especially well given the ability to pause the agent and have end-users authorize or approve specific tools.
🔒Agent Authentication with Arcade AI and LangGraph
One of the main things you do with agents is call tools. One of the hardest parts of calling tools is handling authentication.
You an agent framework that can handle this human-in-the-loop workflow, and technology for handling the auth.
LangGraph and Arcade AI are a perfect combo for this. Arcade AI handles authentication for you, while LangGraph's human-in-the-loop capabilities allow you to easily interrupt an agent to ask for authentication.
YouTube walkthrough: piped.video/Pc9TDng86HU
5
5
20
9,655
We built a LangChain agent that doesn't just chat—it takes real action across tools with proper auth.
Now we want you to outdo us.
Pitch your boldest @langchain × Arcade.dev agent idea by May 14th.
Winner gets 1:1 office hours with our engineers + a full showcase across our platforms.
2
6
18
2,945
While the latest LLM is off writing poetry about cats, your AI app is stuck in read-only mode. But today that changes.
Time to give your AI some hands. 🤖👐
Arcade lets your AI actually DO things on behalf of the user.
Things your agent could never do before:
- Send emails
- Access files
- Update tickets
- Schedule meetings
All with secure auth that doesn't make you want to quit programming.
Built by auth nerds, for AI builders.
Get early access → arcade-ai.com
P.S. Our docs have a higher completion rate than your last D&D campaign 🎲
1
3
16
1,246
🧵 We're featured in @TechCrunch: How we're solving the biggest security challenge for AI agents that actually DO things. A journey from "this is theoretically impossible" to "holy sh*t we're bending reality" 👇
techcrunch.com/2025/03/18/ar…
4
7
17
1,233
Text to SQL 2.0 tutorial shows how to build an AI toolkit that connects your agent to any custom dataset in 30 min or less. No SQL knowledge needed—just ask questions and get data. Complete code + implementation guide at blog.arcade.dev/text-to-sql-…
#AI #SQL #DataAccess
1
1
10
241
Look who just unlocked their second brain 🧠
@TryArcade × @obsdmd is now live. Your AI can finally read, write, and manipulate your knowledge base exactly like you would.
3
11
478
Build this Gmail-Slack AI Assistant in 1 Hour with OpenAI Agents SDK
This tutorial shows how to create a multi-agent system with:
• Multi-turn conversations across services
• Agent handoffs for complex tasks
• Human-in-the-loop approval
• Secure Gmail & Slack integration
Check out the full tutorial: piped.video/OVsjZBmXdbI
#openaisdk #aiagent #buildinpublic
1
2
11
2,505
We built a Telegram bot that schedules Google Calendar events to show a painful truth: auth is killing agent development.
Here's what it takes to make a SINGLE integration work.
Then we rebuilt it with Arcade.dev and Auth code shrunk from 100+ lines to just 9
1
11
775
☁️Last month @CloudFlare casually showed how to build an MCP server on Workers like it's no big deal. Remember when setting up infrastructure was hard? Yeah, me neither anymore.
Their update makes me want to ditch sleep and build agents all weekend. Who's with me? 🙋♀️
#CloudFlareWorkers #MCPServer
blog.cloudflare.com/remote-m…
3
7
1,343
Your AI assistant can now *actually* manage Asana projects.
Not "I'll help you plan" followed by YOU doing all the work.
Arcade.dev's @asana toolkit gives your agent authenticated access to:
- Create tasks
- Manage workspaces
- Update workflows
- Handle attachments
No browser hacks. No auth code.
Build Saturday, demo Monday, blow minds Tuesday.
2
9
402
🚀 The agent ecosystem just hit hyperdrive. This week brought us agent-to-agent communication protocols, auth specs getting real scrutiny, multi-agent systems you can build in less time than it takes to make instant ramen, and @grok joining the API party. Here's your whirlwind tour of what happened while you were busy debugging that one function.
#AIAGENT
1
1
7
751
Developers are building things that make ChatGPT look like a pocket calculator.
Check us out at arcade.dev and join the revolution.
Remember: with great agentic tools comes great responsibility.
6
158
HUGE NEWS: Arcade.dev just landed in the Wall Street Journal! 🔥
The article confirms what we've known all along: authorization is the final boss battle for AI agents. The fancy models? That was the easy part.
Our CEO, Alex Salazar, explained in the piece how we're solving this exact challenge, helping @Shortwave connect their email agent to @NotionHQ —transforming AI from just another chatbot into something that can actually take meaningful action on your behalf.
Massive congrats to our incredible team building the authorization layer that's making the agentic future possible. This is just the beginning.
1
2
6
387
AI agents with admin credentials are like toddlers with flamethrowers—sure they're smart, but do they need THAT much power? Our new guide explains why least privilege isn't just security theater—it's protection against the inevitable "creative interpretations" of your instructions. 🔐 #AIAgents
blog.arcade.dev/sso-for-ai-a…
1
1
6
243
Want to build a Gmail AI agent from scratch? We've got you covered! 🧩 Full tutorial + ALL the code to create an app where users just say "Find emails from Sarah" and it happens.
We guide you through each step on YouTube—from OAuth (simplified by @TryArcade) to deployment.
Start building: piped.video/5fdqz2Xk4dI?si=PLsR…
Code: github.com/ArcadeAI/tutorial…
#buildinpublic
1
6
677
TechCrunch covered our $12M round last week. Now our CEO, Alex Salazar is spilling the tea on how we accidentally built the solution to "the biggest auth problem ever." 🚀
Started building a site reliability agent. Ended up creating something so powerful people thought our demos were fake.
Watch Alex explain how we're busting AI out of chat-mode prison to do something useful in the real world. Because agents that can't access your systems are just fancy autocomplete.
blog.arcade.dev/arcades-12m-…
#techcrunch #funding #startups
3
6
498
🔐 Auth nerds assemble! 🦸♀️ The MCP spec just got a security glow-up with experts from @Microsoft , Arcade.dev, Okta, Stytch, @AnthropicAI and more weighing in.
When this many security folks agree on something, you KNOW it's either brilliant or the apocalypse. Betting on brilliant. #MCPAuth #MCP
github.com/modelcontextproto…
1
5
87
Full Youtube video: piped.video/watch?v=WLb-a2GC…
Open-source Slack agent that connects to @github, Gmail, and Calendar.
Built with @modal_labs + @langchain
Check out the repo: github.com/ArcadeAI/SlackAge…
1
5
1,201
When an agent accesses systems as the authenticated user:
- It can only see YOUR data (not your coworker's embarrassing Slack messages)
- It can only do what YOU have permission for (no nuclear launch codes)
- It creates experiences that feel like science fiction
Critical when dealing with models that occasionally... embellish reality.
1
5
175
More examples
github.com/ArcadeAI/arcade-a…
1
4
1,429
Sign up for the waitlist here:
arcade-ai.com/
1
3
1,040
We've also open-sourced the example here: github.com/ArcadeAI/arcade-t…
See how Arcade.dev cuts through the auth maze so you can focus on building AI that does stuff that matters.
4
180
Seeing builders ship game-changing AI apps without getting bogged down by auth? That's literally why we exist. Thank you for the shoutout @BraceSproul - and even more for showing everyone what's possible when auth just works.
1
2
182
After we rebuilt with Arcade.dev we completely eliminated:
- The Express server for OAuth callbacks
- All Google API client code
- Token management complexity
1
4
40
This wasn't a weekend hackathon project. Our team brings:
- Auth systems expertise from building Okta
- Distributed systems that would make your infrastructure team weep
- Dev experience that feels like magic (but is actually just good engineering)Making the impossible look easy takes work.
A LOT of work.
1
3
47
Our black magic moment: showing ChatGPT draft AND send an email AS YOU, from its context, no copy/paste.
The VCs who saw it converted faster than a Python script to a production incident.
Authentication wasn't just a feature - it was THE feature the entire industry was missing.
1
3
43
The traditional approach is a nightmare:
- OAuth 2.0 implementation
- Token storage/refresh logic
- API endpoint management
All this just to let an AI agent access ONE user service. No wonder <30% of AI projects make it to production.
1
3
51
When GPT-3.5 Turbo dropped function calling, we felt it in our bones: this isn't just another model - it's the new compute platform that all future apps will be built on.
But there was a MASSIVE blocker preventing agents from being useful beyond party tricks with public data...
1
3
61
We set out to build an SRE agent that could:
- Process Datadog alerts at 3am so you don't have to
- Map out diagnostics while you sleep
- Create rule-out lists faster than you can say "have you tried turning it off and on again?"
While everyone was building glorified chatbots, we went for the hard stuff.
1
3
60
The tech landscape was... less than ideal:
- LLMs moved at glacial speed
- Token costs made CFOs cry
- Frameworks were basically beta at best
- Authentication was the final boss with no cheat codes
So we broke ALL the rules and built for "24 months from now" - turns out the future is already here 🤯
1
3
49
The difference? Our agent is now actually useful because it can take real actions, not just chat.
AI without the ability to act on real systems is just fancy autocomplete.
1
3
104
Most "agents" today are just fancy knowledge retrievers - the AI equivalent of "let me Google that for you.
"Real agents need to securely access YOUR systems with YOUR permissions to actually DO things.
The difference between "here's a suggestion" and "I've already handled that for you."
1
3
37
When tech leaders saw our demo, their faces went from 😐 to 😯 to 🤯
"How are you accessing these systems securely?!
"The same way we access the office fridge - with proper authentication and a healthy respect for boundaries.
1
3
41
This isn't just about less code. It's about unlocking what AI can actually DO.
With Arcade.dev, extending this agent to Slack, WhatsApp, or Discord only requires handling the new platform's messaging interface.
Core logic with your LLM? Untouched.
1
3
120
We showed early builds to Harrison at @langchain , and he got it immediately.
Now we're building "first of a kind" sample apps together that make current solutions look like they're running on Windows 95.
This is what happens when two teams obsess over developer experience.
1
3
40
The rules are dead simple:
- One idea per person
- Must use LangChain + at least one Arcade.dev tool
- Must be buildable in a weekend
- Must solve a real problem with authenticated actions
Need inspiration? @torresmateo built this social media agent with Notion in 30 minutes ⬇️
piped.video/Uq7NqoUrlTc
1
3
80
Want to try the code yourself? Watch the full video here: piped.video/6pDyjopGW7E
1
3
295
There's nothing better than seeing teams build cool things with our product. This agent is awesome. Already playing with it for our team.
3
763
To enter:
1. Create an Arcade.dev account at account.arcade.dev/register
2. Submit your concept here: try.arcade.dev/m/langchain_c…
3. Join our Discord where finalists will get feedback directly from our engineers
Submission window closes May 14th at 11:59 PM PT.
1
2
308
Finally, someone made Twitter API access not suck. Love how you combined RAG + memory here - clean architecture, no BS. @TryArcade toolkit making it dead simple to build AI clones is chef's kiss 🚀
Might create an AI version of myself that actually writes good tweets.
1
3
574
Check out the repo: github.com/spartee/arcade-ob…
2
205
Sharp eye on the docs! While we do make LLMs play nice with tools, you're not locked into our completion endpoint. Can just call the tools directly.
Your agent, your rules.
docs.arcade-ai.com/home/use-…
1
1
62
We're looking for agents that actually DO things. Not just chat, not just retrieve—agents that take real actions on your behalf through secure authentication.
Create a social media manager, a content scheduler, a research assistant that updates your docs... possibilities are endless.
1
2
105
+ If you're attending the LangChain Interrupt conference this week, swing by our booth to discuss your agent idea with our team!
2
245
🔥 @Grok v3 models are now API-accessible and developers are already going wild.
xAI's spicy contender brings its unique flavor to the dev toolkit—less filtered, more creative, and surprisingly capable. Time to see what happens when you let Grok loose on your next project.
Prediction: chaos, but the productive kind. #GrokAPI #AIDevAccess
docs.x.ai/docs/overview
1
1
320
If you want to start playing with the new streamable HTTP transport for MCP, there's a live demo here:
blog.arcade.dev/announcing-n…
1
316
👀 Look who else is building the robot economy! Grab early access and our SDK could help make @trySkyfire integration smoother than Quicksort algorithm on sorted data. Because the only thing better than AI agents with hands is AI agents with a wallet. 🤝💵
1
59
@trung_rta While everyone's waiting for AI to make money, you're already having it ship products. 🚀
Ready to upgrade that "$500k consultant" from "strategy deck" to "actual deploy"? That's what Arcade's built for.
1
1
32
⚡️@TryArcade just showed how to build a useful multi-agent system in under an hour using @OpenAI Agents SDK.
Remember when building useful AI systems took days and a PhD? Now it takes less time than your last stand-up meeting.
The future is lazy (in the best possible way). #AgentsSDK
piped.video/OVsjZBmXdbI
1
1
70
@nbarbettini , We knew you were our kind of wizard when you got as excited about OAuth flows as we did about AI capabilities. 🧙♂️ Thanks for helping us turn "wouldn't it be cool if..." into "holy shit, it actually works."
1
52
If you want to start to play with the streamable HTTP transport, we just spun up a live demo.
blog.arcade.dev/announcing-n…
11