


Co-founder @keycardai. 🛠️ @passportjs. ❤️ HTTP 401-403. #openid #oauth #wimse #mcp #x402


Agents. Humans. Same.
the most import skill in 2026 is still being able to tell your agents "wtf is this crap, no, not like this" if you stop doing that, you're basically fucked, despite what some devfluencers would like you to think.

the most import skill in 2026 is still being able to tell your agents "wtf is this crap, no, not like this" if you stop doing that, you're basically fucked, despite what some devfluencers would like you to think.

I see a lot of people describe their agent setups. But I don’t see many ways for other people to quickly clone and run that setup. The distribution model is lacking. What projects do you know that are solving this?

This is the way. People are shocked that I use AI all the time, but never on my own machine. Open claw, Hermes, grok @bot all running in a Mac mini. Separate email. Separate phone number. Separate Twitter (@clawdreyhepburn if you want to say hi). Don’t let agents impersonate you - give them their own identities.
When you collaborate with coworkers, you are not delegating your access to them. Yet this is how most AI access works. That’s going to change. Collaboration, not delegation.

When you collaborate with coworkers, you are not delegating your access to them. Yet this is how most AI access works. That’s going to change. Collaboration, not delegation.

Skills over MCP are great. But seeing skill:// as a URL makes me look side eyed at the protocol itself.

Agents aren’t infrastructure

Personal agents demand this, since the services they call don’t control the network or the harness. Yet they need to be able to challenge for credentials.

The trend to move things out of transport layer to app layer is clear. Agents are no different.

Yes, credentials themselves need to be kept out of LLM context - but that’s doable without resorting to extremes. DPoP in particular helps here.

Approaches like credential injection and proxying limit the agent’s ability to reason.

Credentials and corresponding challenges need to be in-path of a harness so that an agent can reason over its own access paths. Challenges in particular are useful here.

This is a one-shot prompt, sent to devs on X rather than an LLM. It’ll exist shortly by virtue of tweeting it.
MCP vs CLI for LLMs is the wrong discussion. People are debating at the wrong layer. Both work incredibly well as long as they are run through a repl like environment that can persist state. The funny thing about CLI is that CLI tools are usually accessed through BASH which happens to be a repl with persistent state (the file system), therefore cli works much better than mcp. But this isn't even close to being an intrinsic property of MCP. Just means that we need better harnesses. Right now the best repl for this are: - bash + fs - jupyter kernels - codemode type repls, usually quickjs LLMs understand the concept of forward evolving a system to solve a need very well. This comes from the agentic RL. They understand how to change the state of a codebase or acquire data from databases and then operate on it similarly to humans. But without an execution environment, they actually cannot do this properly. My bet? Sometime soon someone will (or has already?) create a embeddable, sqlite style mini execution environment that parses bash, typescript, or tool calls into a common IL execution plan that's easy to security check before executing. And design this specifically for durable execution environments. Then we will just connect cli, mcp, webmcp, whatever to that and it accepts any of the input modalities as they can all be represented as each other.

insane 48 hours keeps getting worse... 🤯 banks finding unauthorized AI agents inside customer accounts, some acting on their own, and execs can't tell human from agent or authorized from rogue every agent needs identity + intent, dynamic, at runtime, scoped to the task and gone the second it's done
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios. The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios. Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said. Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior. Read my latest for Axios here: axios.com/2026/10/22/ai-agen…

We've spent a lot of time at @KeycardAI thinking about how authentication and authorization need to evolve as agents take on more work. API keys and OAuth scopes were never designed to account for tasks, delegation between agents, or agents acting on their own. I wrote about how we're approaching this and what we think needs to change. keycard.ai/blog/defining-aut…