
AI agent infra @TryArcade | NOT 7 (Declarative agent engine) | MCP, Agent architecture, Enterprise AI
California, USA
Joined May 2014
- Tweets2.7K
- Following531
- Followers277
- Likes5K
Your agent's biggest token cost is not the conversation. It is the tools.
Context bloat due to tools needs to be talked about far more than it is. Every turn, an agent pays to declare its tools, select from them, and call them. At thousands of tools, the definitions alone run past a million tokens. It is an underestimated cost problem, and at scale it will bite organizations hard.
The research behind this is timely. It proposes a different approach entirely: stop listing tools to the model. Treat selection as recommendation. The agent asks for what a task needs and gets only that, when it needs it. Measured end to end: the bill dropped by up to 40%, and coverage of every tool a task actually needs rose 57% over the standard baseline.
Missing one required tool can derail a multi-step agent task.
@avoguru explains how agents can find the right tools among thousands, how to evaluate retrieval accuracy, and when a fixed list works better.
aaif.io/blog/how-to-give-an-…
Excited to announce Enterprise MCP Hosting with Arcade Deploy.
We’ve built thousands of tools. But your business needs tools only you can write.
Tool Patterns and Toolbench bring our tool-engineering experience to your developers. Arcade Deploy brings those tools into production.
Run arcade deploy. Your custom MCP server inherits your Arcade environment’s configured identity, authorization, policy enforcement, and audit logging.
Our direction is simple: give you the tools to build the way we build, with the runtime controls to run what you create.
Here is a demo: youtube.com/watch?v=1IQW8IjS…
Guru retweeted
Arcade.dev's ToolBench now covers 90,000+ open-source MCP servers and over 600K agent-ready tools, making it the most comprehensive look at the MCP server and agent tool ecosystem out there.
Across all of it, growth is far outpacing safety and quality, with only 260 servers earning an A and the riskiest tools getting the least documentation. arcade.dev/blog/toolbench-re…
Guru retweeted
Guardrails stop a rogue agent most of the time. Probably. Hopefully.
"Probably" doesn't cut it in an enterprise. You need to know your agent won't send that payment, delete that record, or fire off that email. That's what AI agent governance is for.
We put together videos, articles, and a downloadable playbook to help: arcade.dev/aiagentgovernance…
🐶 (paid in treats, not dollars)
Guru retweeted
Execution Tool Logs are now live in Arcade. Every tool your agents run traced: which tool, which user it ran for, every retry, and the error if it failed.
Arcade already decides what your agents are allowed to do. This is the record of what they actually did.
@avoguru and Jose Barrueta wrote it up: arcade.dev/blog/execution-to…
Good guidance
I just published a manifesto for all the developers out there who use an AI coding tool, but feel like they're not shipping much faster
I wrote 10 principles for changing how you build software with AI, based on what I've seen work for teams across Amazon
kiro.dev/topics/frontier-eng…
1
59
Guru retweeted
With people, the actor and the accountable party are the same. The analyst who runs the report answers for it, which is why “is this user allowed?” has been enough.
With agents, that framework comes apart. The agent acts, a human carries the consequence, and the controls you already run aren’t built to handle that new reality.
@torresmateo on why: arcade.dev/blog/ai-agent-gov…
Guru retweeted
Got the whole team into SF to celebrate @TryArcade's Series A.
We started Arcade before most people believed agents were going to matter, and spent a stretch explaining a problem only those on the bleeding edge had run into yet. Now, some of the largest enterprises in the world are moving agents into production and the round is what lets us increase the gap between where we are, and everyone else.
Standing in that room last week, I feel good about the future because I was surrounded by extraordinarily talented people who will get us there.
RT @TryArcade: Last week, we brought the whole team into SF to celebrate our Series A.
We raised this round to make Arcade the default ans…
Guru retweeted
NVIDIA just launched the Open Secure AI Alliance, and Arcade.dev is a founding member.
The idea behind the alliance aligns with our core premise that securing AI is too big for any one vendor, and defenders need open tools they can trust.
Our part is the actions runtime: it enforces who authorized each action, executes it reliably, and governs every agent from one place.
Better models don't get agents into production. The security layer underneath does.
Here’s why we joined: arcade.dev/blog/arcade-joins…
Over the past ten months I have spent most of my time on agentic development inside several organisations. The technology side has genuinely got easier. Models improved, harnesses matured, the protocol conversation settled down. What has not improved is the room where these platforms get designed.
gnanaguru.com/blog/a-room-fu…
🤖 Made with AI
Guru retweeted
Replying to @Forbes
@Forbes on our @SmitheryDotAI acquisition: “...the first time an MCP registry with real developer mindshare has been bought by the layer that governs execution.”
Enterprises moving agents into production need both a catalog of tools and a runtime that can authorize and audit them. Those two things are merging into one product.
Full feature: forbes.com/sites/janakiramms…
Guru retweeted
Joining @justheena and @msft4startups for #FounderFriday on 8/25.
Talking about why we held off building until we had a real demand signal, how the market pulled us toward enterprise, and why the workflows that pay off are the ones agents can finish rather than assist with.
Register: linkedin.com/events/74915328…
Guru retweeted
Smithery has been acquired by Arcade!
It’s been an incredible year building Smithery alongside the MCP community. I'm proud that we've built something that helped process millions of tool calls and made 10K+ MCPs from developers around the world more discoverable.
My co-founder @kamathematic is joining Arcade to keep building toward our vision of making tool calls the new clicks.
As for me, I’ll be taking some time to explore what’s next. More soon!
Replying to @SmitheryDotAI
@SmitheryDotAI, the leading public registry and hosting platform for MCP, is now part of Arcade.dev.
Smithery made finding and running an MCP server take one click. Pairing that developer experience with the security, reliability, and governance that Arcade is known for, gives enterprises both to lead the next generation of agentic AI.
Smithery co-founder @kamathematic is joining us to keep building it.
More from @TheMostlyGreat: arcade.dev/blog/smithery-joi…
Guru retweeted
Replying to @SmitheryDotAI
@SmitheryDotAI, the leading public registry and hosting platform for MCP, is now part of Arcade.dev.
MCP adoption has outpaced tool quality. As enterprises put production agents on it, quality and trust become the priority. Smithery and Arcade together bring both.
Glad to welcome @kamathematic.
More: arcade.dev/blog/smithery-joi…
Super excited by this announcement, and welcome again @kamathematic to the team.
With one of the best minds in the industry now on board, Arcade brings together the two things the agentic internet has been missing: one click to find and run any MCP server, and the trust to put it into production.
Replying to @SmitheryDotAI
@SmitheryDotAI, the leading public registry and hosting platform for MCP, is now part of Arcade.dev.
Smithery made finding and running an MCP server take one click. Pairing that developer experience with the security, reliability, and governance that Arcade is known for, gives enterprises both to lead the next generation of agentic AI.
Smithery co-founder @kamathematic is joining us to keep building it.
More from @TheMostlyGreat: arcade.dev/blog/smithery-joi…
Guru retweeted
Just had a workshop at the Venetian where I talked about background agents with @TryArcade and @daytonaio at @Ai4Conferences
Come say hi at our booth tomorrow if you’re here for ai4 and building enterprise agents!
Trying to run too many agents, kept losing track of which are still going and what the status was.
notch-mcp gets you live updates direct from your agents, at the top of your Mac.
It's an MCP server, so it works from Claude Code, Cursor, Windsurf, Zed instead of being tied to one of them.
Contributions welcome. If you want to make it a lot cooler, just submit a PR.
github.com/avoguru/notch-mcp
32























