EDAMAME Security
On each machine · freeThe app for macOS, Windows, Linux, iOS and Android. It hardens the machine and watches its AI agents from outside the agent.
Get startedSecurity That Moves at the Speed of Code
Software is now built by humans, pipelines and AI agents, all touching code, secrets, production and other critical company resources. EDAMAME proves that each machine and each AI agent can be trusted, by its security posture and its runtime behavior, and binds access to that proof: EDAMAME Security on workstations, EDAMAME Posture on CI/CD runners and any server running AI agents, cloud or self-hosted, and EDAMAME Hub for the fleet.
Meet our team
The engineers, researchers and advisors behind EDAMAME.
Our story
We build the trust layer for AI agents and the machines they run on.
EDAMAME Technologies is a French cybersecurity company founded by Frank Lyonnet, PhD, a former INRIA researcher and serial cybersecurity entrepreneur. Our products prove that each machine and each AI agent can be trusted, by its security posture and its runtime behavior, then bind access to that proof.
EDAMAME Posture
On runners and serversThe CLI for CI/CD runners and any server running AI agents, cloud or self-hosted.
CI/CD securityEDAMAME Hub
For your teamOne view of every workstation, CI/CD runner and AI agent, with policies, conditional access and audit evidence. Devices report; the Hub never takes control of them.
Try EDAMAME HubEDAMAME Portal
Your accountWhere you sign in from the app, and where the Cloud LLM plan behind the AI Assistant, your devices and your API keys live.
Open EDAMAME PortalHow we build EDAMAME
Four rules every product follows.
Developers, contractors and AI agents keep moving; the company gets proof. Nothing in EDAMAME takes a machine away from the person who uses it.
Reporting-only
No one can remotely control, wipe or run commands on a device through EDAMAME. Fixes run locally, and can be undone wherever a rollback exists.
Outside the agent
EDAMAME watches AI agents through the operating system's own observation interfaces, with no plugin inside the agent.
Consent first
When a device joins EDAMAME Hub, a consent screen lists exactly which information will be shared.
A human decides
The AI Security Assistant reports its actions and escalates risky decisions to a human.
Public proof
Patented, published and open to review.
FR3146742
Our decentralized posture certification is protected by French patent FR3146742, granted in April 2026, with international applications pending. It certifies the posture of an unmanaged device without the verifier accessing or controlling it.
LISTIC · CNRS
Runtime verification for AI agents is grounded in research with the LISTIC laboratory (CNRS): "Runtime Security for Agentic Systems", in open access on HAL.
Read the paper on HALApache 2.0
The security checks are public threat models that anyone can review, and their updates are cryptographically signed. The EDAMAME Posture CLI, the helper and the CI/CD actions are open source too.
Threat models on GitHubMilestones
From posture proof to AI agent trust.
- Sep 2022
EDAMAME Technologies is founded in France.
- Mar 2023
Patent application filed for decentralized posture certification.
- Apr 2026
French patent FR3146742 is granted.
Headquartered in Paris with operations in San Francisco.