Olame_Elam🇨🇩

Building the accountability infrastructure the agent economy runs on. Runtime authorization for autonomous agents before it executes

[Olame_Elam🇨🇩 @Olame_M]

Currently building the infrastructure layer nobody had built yet.
Autonomous agents are running in production systems right now.
Valid credentials. Clean individual requests.
No layer governing what they actually do with those credentials.
#AgenticAI #AIGovernance #LangGraph

[Olame_Elam🇨🇩 @Olame_M]

That gap is where the next class of enterprise breach lives.

AgentGate closes it — runtime authorization that watches the sequence,not just the request. Stateful. Pre-execution. Open source.
3 lines to integrate.

→ https://t.co/8oWbizExaT

#AgenticAI #AIGovernance #LangGraph

[Olame_Elam🇨🇩 @Olame_M]

Currently building the infrastructure layer nobody had built yet.
Autonomous agents are running in production systems right now.
Valid credentials. Clean individual requests.
No layer governing what they actually do with those credentials.
#AgenticAI #AIGovernance #LangGraph

[Olame_Elam🇨🇩 @Olame_M]

If this is useful, a ⭐ on GitHub helps more than you think.

[Olame_Elam🇨🇩 @Olame_M]

🧵 Your AI agent just read 847 files in 4 minutes.

Your logs show 847 authorized requests.

You have no idea what's about to happen next.
#AISecurity #AgenticAI #CyberSecurity

[Olame_Elam🇨🇩 @Olame_M]

AgentGate is the pre-execution authorization layer that tracks behavioral patterns across 24h sessions and blocks kill chains before they fire.

Open source. MIT. Three lines to integrate.
@owasp
Live demo: https://t.co/5A5cnYWtrw
GitHub: https://t.co/8oWbizExaT

[Olame_Elam🇨🇩 @Olame_M]

@Threads If you're building with @langchain or working on agent security, I would love your feedback.

Built this with @owasp's Agentic Top 10 in mind — most of the gaps they identified (goal hijacking, delegation abuse, excessive agency) are exactly what AgentGate addresses.

[Olame_Elam🇨🇩 @Olame_M]

@Threads AI agents with valid OAuth tokens can still:
- read files outside their scope
- be delegated more perms than their parent had
- exfiltrate data below rate-limit thresholds
- get hijacked via a doc they're processing

OAuth checks *who you are*.

[Olame_Elam🇨🇩 @Olame_M]

This doesn't exist anywhere else I've seen.
MIT license, works with @langchain, AutoGen, or any custom agent.

pip install agentgate-pdp

GitHub: https://t.co/yuirqzQmGf

[Olame_Elam🇨🇩 @Olame_M]

Decision: PERMIT, ESCALATE, or DENY in <100ms

The delegation chain check is the part I'm most proud of.

Child agents can NEVER exceed what their parent was authorized to do.
Every hop attenuates scope. Orphan delegations (parent not in registry) = hard deny.

[Olame_Elam🇨🇩 @Olame_M]

@yomiekila ahahahaha tu parles mal😭😭

[Olame_Elam🇨🇩 @Olame_M]

@yomiekila un peu douteux

[Olame_Elam🇨🇩 @Olame_M]

@JoanLaportaFCB chef tout chez toi

[Olame_Elam🇨🇩 @Olame_M]

@yomiekila si ça me donne les papiers, #letsgo

[𝓐 @Sincerely_AO]

Building a habit of reading your bible and praying for at least 30 minutes a day will change your heart, mindset, and relationship with God for the better. Highly recommend.

[… @Unkle_K]

Well… some of us weren’t stupid

[Olame_Elam🇨🇩 @Olame_M]

@AmazingBua Tu veux tadah😭😭😭