Pinned
One platform to connect, secure, control, and observe every AI agent across your org.
This is the AI control plane.
00:00
Speakeasy
1,120 posts
Speakeasy
@speakeasydev
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.
- Speakeasy reposted
We wrote up how MCP servers are being used to attack the agents that connect to them. An agent trusts an MCP server to describe its own tools and return its own results. The exploits that abuse that trust have names now: tool poisoning, rug pulls, cross-server shadowing. None - Speakeasy reposted
Today we are shipping the Interactions API into GA, our new API designed to let you orchestrate across models and agents in the same interface, and is our new default API going forward. Interactions sets the stage for the new era of Agents : ) - Speakeasy repostedThere's plenty written about what to govern in AI systems and almost nothing about how to actually enforce it on live traffic. We wrote up how we do it. The most important place to enforce a policy is the tool call, because that's where an agent's intent turns into a real
- Speakeasy repostedOur AI bill spiked May 22nd to May 24th. A single session to refactor a codebase cost us $2,000 dollars. Amazing what you can do with the data in our product now to pin point AI usage across agents, mcp and skills. @da_adler caught in the act :DYou shouldn't have to wait for the invoice to find out what your AI cost. Speakeasy sits in the path of every AI call, so spend is managed in real time, not reconciled after the fact. Personal licenses, enterprise seats, provider APIs, and your model router report into one
- Blocking MCP servers doesn't stop shadow AI. It just moves it to personal laptops. The fix is to make the sanctioned path the easy one. With MCP authorization you can define roles scoped to specific people, teams, and MCP servers. Paired with the Speakeasy registry, IT can stand
00:00 - Speakeasy repostedBoth frontier labs filed to go public this month, and their pricing pages already said what the prospectuses will: the subsidized era of AI is over. Every new model used to launch cheaper than the one before it. GPT-5.5 launched at double the price of GPT-5.4, and GitHub
- Speakeasy repostedYou shouldn't have to wait for the invoice to find out what your AI cost. Speakeasy sits in the path of every AI call, so spend is managed in real time, not reconciled after the fact. Personal licenses, enterprise seats, provider APIs, and your model router report into one
- You shouldn't have to wait for the invoice to find out what your AI cost. Speakeasy sits in the path of every AI call, so spend is managed in real time, not reconciled after the fact. Personal licenses, enterprise seats, provider APIs, and your model router report into one
- Speakeasy repostedWhen Mythos leaked in April, cybersecurity stocks like Akamai fell 20%. The market's verdict was that frontier AI means attackers win. Six weeks later, defenders using the same model found more than 10,000 high and critical vulnerabilities in a month, including 2,000 at
- Speakeasy repostedThe release and near-immediate unrelease of Fable put enterprise AI spend back in the headlines this week. Good moment to resurface something we wrote a while back, because one of the quieter drivers of that spend is MCP. Every tool an MCP server exposes has to be described to
- Speakeasy repostedThe official MCP server from @datadoghq exposes 142 tools across 22 toolsets. Connect it to an agent and the context window fills up with tool definitions before the agent does a single useful thing. Beyond the context, every irrelevant tool is one more thing the model has to
- Speakeasy repostedHow many agents are running in your org right now that you don't know about? A Cloud Security Alliance survey puts the odds at 82% that the answer is at least one. If you don't even know they exist, you also don't know what they're doing. What tools they're calling. What data
