Your AI control plane. Safely scale AI across your org. Connect, secure, and monitor AI in real time. Every MCP, skill, CLI and agent session governed.
SF & London 🍻
Joined May 2022
- Tweets1.3K
- Following71
- Followers1.6K
- Likes1.4K
Speakeasy retweeted
Every company we talk to has shadow AI running on someone's laptop. If you manage your Macs through @JamfSoftware, adding @speakeasydev lets you detect and govern every AI tool on them.
Check us out: marketplace.jamf.com/details…
Speakeasy is now on the @JamfSoftware Marketplace. Deploy the Speakeasy Device Agent through Jamf Pro to govern every AI agent on your managed Macs.
marketplace.jamf.com/details…
We'll be at @AgenticAIFdn's AGNTCon + MCPCon this week in San Jose. Come say hi.
Thank you to everyone who joined us and @Tailscale at The Allowlist to close out @aiDotEngineer NYC.
Put your MCP servers behind your Tailnet, then secure the rest of your enterprise AI usage with Speakeasy.
Speakeasy retweeted
Good crowd at The Allowlist last night. Thanks to everyone who came and to @Tailscale for co-hosting.
If you missed it, the @speakeasydev MCP gateway now runs on your tailnet with the public route shut.
speakeasy.com/blog/release-p…
Speakeasy retweeted
If your employees need Tailscale to reach internal systems, your agents should too.
We've partnered with @Tailscale so the @speakeasydev MCP gateway runs on your tailnet, with the public route closed and every tool call tied to a user and device.
speakeasy.com/blog/release-p…
We're partnering with @Tailscale to take enterprise MCP servers off the public internet.
The Speakeasy MCP gateway runs as a node on your tailnet, the public route closes, and every tool call carries Tailscale user and device identity.
speakeasy.com/blog/release-p…
Speakeasy retweeted
Last time i was at an @aiDotEngineer conference this graph was flat. The daily growth of MCP calls run through our platform is rapidly accelerating. Enterprise adoption of MCP is very underestimated. I expect 10 million a day in just a few weeks and 100 million a day early next year.
Come by @speakeasydev booth if you want to securely scale MCP use at your company.
Speakeasy retweeted
Put our @linear agents on the golden path today by having them connect to our engineering MCPs through @speakeasydev's MCP gateway 😎.
Security ✅
Observability ✅
Identity ✅
My CISO can sleep easy now.
Speakeasy retweeted
If you want to make sure all your MCP connections regardless of harness are secured behind Tailscale check out @speakeasydev
Speakeasy retweeted
If you're in NYC next week, come chat with my CEO.
P.S: He lovessss talking about agent identity & autonomous agents. :)
Catch Sagar at @aiDotEngineer NYC next week.
"Permission to Act: Giving Agents Authority in Regulated Industries," Wed at 2:55pm ET.
Then come find us at the booth.
Speakeasy retweeted
Claire's dashboard is exactly what enterprises need. Our customers have been getting this for months. Seeing all of their usage, costs, roi and expensive sessions all in one place and accessible through mcp. The best part is everything is tied to their Okta/IDP groups so you don't have to the manual work of figuring out how this maps to budgets and finance 💸
I can tell you where every penny of token spend is going against my business, can you?
This whole analysis cost me $0.12
Here's how I track everything my agents are doing to tie AI use to business impact 🧵
Catch Sagar at @aiDotEngineer NYC next week.
"Permission to Act: Giving Agents Authority in Regulated Industries," Wed at 2:55pm ET.
Then come find us at the booth.
Speakeasy retweeted
We're hosting a happy hour with @Tailscale in New York on October 14. The Allowlist.
Most MCP traffic is internal, but the servers are exposed to from the public internet. Why? Speakeasy makes it trivially easy to move all your servers to your Tailnet. Put them on your private network and improve the security of your AI workflows in seconds.
If you're in New York for @aiDotEngineer, come join. Drinks on us.
luma.com/po85yuqq
We're hosting a happy hour with @Tailscale in New York on October 14. The Allowlist.
Most MCP traffic is internal, but the servers are exposed to from the public internet. Why? Speakeasy makes it trivially easy to move all your servers to your Tailnet. Put them on your private network and improve the security of your AI workflows in seconds.
If you're in New York for @aiDotEngineer, come join. Drinks on us.
luma.com/po85yuqq
Speakeasy retweeted
@CoinbasePltfrm shipped a global kill switch for their AI agents. We built one of those. Nobody ever used it.
If your only control takes down every agent in the company, using it is an outage. The downsides of getting it wrong are high so likely no one will ever use it.
Second time around we rebuilt it around a smaller question: can I pause one person's tool calls in ten seconds and undo it in ten minutes?
What we focused on:
- Scope is identity + servers. Pick the person, pick all servers or a named few. "All" means all, including ones registered after the pause starts. We froze that list in v1 and it's a trap: someone onboards a server mid-incident and it's silently uncovered.
- All checks run in the gateway, before the server instead of relying purely on agent hooks.
- Check can't complete, call is denied but customers can choose to fail open depending on their tolerance.
- Session stays up. Forbidden on tool calls with a code and a reason. Otherwise agents hammer reconnect and the human thinks it's a network fault.
- It expires. Anything needing a human to remember to undo it becomes a permanent deny someone debugs in six weeks.
So now it is cheap to pull, cheap to undo, so it gets pulled while the problem's small. Same reason incident.io works: declaring an incident has to be cheap or nobody declares one.
Ends up looking like access control, not circuit breakers.
Check out our kill switch feature for agents at @speakeasydev speakeasy.com/docs/ai-contro…
Cloudflare open sourced their SDK generation pipeline.
Cloudflare launched MCP Portals, their MCP gateway.
Cloudflare launched a model fine tuning platform.
Therefore @speakeasydev will launch a model fine tuning platform 😌
Speakeasy retweeted
800 MCP servers and 5,000 tools behind one internal gateway. That's the number Uber published this week for its own agent infrastructure, and it says something sharper than another "we adopted MCP" post: tool sprawl inside a single company can now rival the size of a public API ecosystem.
The mechanism is unglamorous, and that's the point. A registry acts as the control plane, tracking ownership and which tools are enabled for which agents. A proxy gateway sits in the data plane, translating MCP calls into whatever the backend actually speaks, HTTP, gRPC, TChannel, routed through Uber's existing Muttley service mesh. Authorization happens per tool, not per service, and PII gets redacted from responses before an agent ever sees it. An AutoCrawler component pulls API definitions straight from Uber's IDL registries so teams don't have to hand write MCP wrappers for things that already exist.
Uber isn't alone in deciding MCP needed a front door. Cloudflare published its own enterprise reference architecture the same season, built around a different bet: distribute MCP servers globally instead of centralizing them, and fix the token bill before the security bill. Its Code Mode approach claims a 94% cut in context consumption, from roughly 9,400 tokens down to 600 for the same 52 tools, by making agents discover tools progressively instead of dumping every schema up front. Speakeasy, MintMCP and a few others are now publishing straight up buyer's guides for "best MCP gateway," which is usually the tell that a pattern has become a market.
What nobody's pricing in yet is the irony sitting underneath all of it. MCP got popular because it let any team stand up a tool in an afternoon without asking platform engineering for anything. The gateway is the bill for that freedom: a whole second system whose entire job is reintroducing the governance everyone skipped the first time.
My read: this is where agent infrastructure spending actually goes in year two. Not more model calls, the plumbing that makes five thousand tools safe to point an agent at.
xitter.cf/UberEng/status/2106072…
What does it take to scale MCP across Uber?
Our latest engineering deep dive traces Uber’s MCP journey. From early, community-driven adoption to building MCP Gateway, a centralized platform designed to make AI-powered developer workflows easier to discover, manage, and scale.
The team dives into how MCP Gateway brings together discovery, control, and proxy layers to connect AI agents with Uber’s existing backend systems.
Featuring work from several authors recently recognized in the @AgenticAIFdn newsletter, the team unpacks the architecture, challenges, and lessons behind building MCP Gateway.
Speakeasy retweeted
Private drinks with the private network. Join the @Speakeasydev and @Tailscale teams at a real life speakeasy. It's on the last day of @aiDotEngineer, we'll be chatting about privacy and security in the age of agents.
Space is limited: luma.com/po85yuqq
We're teaming up with @Tailscale to keep your MCP connections off the public internet.
Join us at a real speakeasy on the last day of AIE NYC, to chat about privacy in the age of AI.
Space is limited.
