Stephane Nangue

Sous la complexité apparente de la réalité, réside une simplicité déconcertante.

[Stephane Nangue @stephanenangue]

@kubesploit I have added a helm chart that could help you deploy warden easily on kubernetes : https://t.co/hGl00xDS03

[Stephane Nangue @stephanenangue]

Every single company that wants to connect AI agents to the external world will implement an egress proxy that broker credentials. Since last year I have been working on an identity-aware egress proxy that treats AI agents a first class citizens : https://t.co/2y0A1iiva5

[LangChain @LangChain]

Introducing the sandbox Auth Proxy: A way to control the boundary between agent-generated behavior and the rest of the world.

An explainer from @hwchase17

[Stephane Nangue @stephanenangue]

@LangChain @hwchase17 Every single company that wants to connect AI agents to the external world will implement an egress proxy that broker credentials. Since last year I have been working on an identity-aware egress proxy that treats AI agents a first class citizens : https://t.co/2y0A1iiva5

[Stephane Nangue @stephanenangue]

@QuinnyPig Very good thread. I will love to hear your feedback about this : https://t.co/gOPoDU9Wfn

[Stephane Nangue @stephanenangue]

I agree with this, agents should have their own identity

[Corey Quinn @QuinnyPig]

4. Agents need their own identity. Today every action is laundered through the human's IAM role, so the audit log reads "corey@duckbill did this" when the truth is "Claude's third retry at 2am did this."
First-class agent identities: scoped, attestable, time-limited, revocable.

[Stephane Nangue @stephanenangue]

We have been building Warden, a tool that connects AI agents to enterprise systems. Agents send requests using their identity (JWT, Cert with SVID support) and Warden does not only credential brokering, but also credential minting. https://t.co/gOPoDU9Wfn

[Corey Quinn @QuinnyPig]

2. Stop making me fish for API keys every time the agent wants to light up a new service. The fix is secrets brokering: the platform holds the secret, the agent gets a handle, calls go through. A compromised agent can't exfiltrate what it never had.

[Corey Quinn @QuinnyPig]

2. Stop making me fish for API keys every time the agent wants to light up a new service. The fix is secrets brokering: the platform holds the secret, the agent gets a handle, calls go through. A compromised agent can't exfiltrate what it never had.

[Stephane Nangue @stephanenangue]

@ClaudeDevs Good news. I cancel my Max x5 plan in June and switch to Codex.

[Stephane Nangue @stephanenangue]

I am looking for contributors.

[Kubesploit @kubesploit]

Warden is an open source runtime access gateway that lets AI agents, pods, pipelines, and services use identity-based policies to reach cloud APIs, databases, and storage without storing long-lived credentials

➜ https://t.co/UBTJVR2DJL

[Kubesploit @kubesploit]

Warden is an open source runtime access gateway that lets AI agents, pods, pipelines, and services use identity-based policies to reach cloud APIs, databases, and storage without storing long-lived credentials

➜ https://t.co/UBTJVR2DJL

[Stephane Nangue @stephanenangue]

@clare_liguori Completely unnecessary because the CLI can do the work. This is typical the hyperscaler strategy : own everything so that we can collect telemetry and eventually charge the customer for agent usage.

[Stephane Nangue @stephanenangue]

@tpschmidt_ I tried it out and it is pretty good. But the fact that the Agent still hold credentials really annoys me. That why I am building a tool like Warden (https://t.co/gOPoDU9Wfn), which supports very rich policies for external systems.

[Stephane Nangue @stephanenangue]

@akshay_pachaar Great article. You could have included Goose, one of the only open source Agent Harness backed by AAIF.

[DeepSeek @deepseek_ai]

🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.

🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.

Try it now at https://t.co/GCdiMzk1Dl via Expert Mode / Instant Mode. API is updated & available today!

📄 Tech Report: https://t.co/drlDrxkYtp
🤗 Open Weights: https://t.co/T13Y8i7SDM

1/n

[Marc Benioff @Benioff]

No browser required. Our API is the UI. 🔓 Salesforce Headless 360 just exposed our entire platform — apps, workflows, metadata, Agentforce & Slack — as unified APIs, MCP tools & CLI. Build on any surface. Give Agentforce deep, trusted context. Stop just using Salesforce. Start building with it. 🚀
https://t.co/8po9KM75RH

[Stephane Nangue @stephanenangue]

@dangtony98 @matsiiako I have been working on a similar opensource project that relies on agent having real identity (JWT, Certificate and SPIFFE) and that injects credentials into any API including AWS with SigV4 : https://t.co/gOPoDU9Wfn

[Matthew Slipper @mslipper]

IMO putting secrets in env vars is an antipattern. The Vercel thing makes that clear. Workloads shouldn't have secrets at all. Keep them in a vault instead and have an egress proxy inject them on the way out. This will seem obvious in a year.

[Stephane Nangue @stephanenangue]

@konstiwohlwend <1h is impractical in real life. And even if app manage to use that short-lived JWT, with AI augmentation it can still stolen and exploited within it lifetime. Federation is not the fix, it is just a patch.

[Stephane Nangue @stephanenangue]

@WallisDev We need to move on from secrets, static and dynamic, short-lived and long-lived.

[Stephane Nangue @stephanenangue]

@garrytan MCP will not survive enterprise deployment and scale. The fact that it is stateful is one of its main drawbacks