This is "verify me" AI — not "trust me" AI.
CIRIS Agent is a personal AI assistant where every action passes through an 11-step ethical reasoning pipeline before execution. Four independent evaluations: common sense, domain-specific, principled, and intuition, plus a four-faculty conscience validate every decision. If the system isn't sure, it escalates to you. You always have final authority.
Every decision is cryptographically signed. CIRISVerify anchors your agent's identity in your device's Secure Enclave and signs actions with Ed25519 across five trust levels. The audit chain is hash-linked and tamper-evident. You can trace any reasoning chain and prove what happened.
Security that gets stronger over time. The Coherence Ratchet is our attempt to make honesty compound and deception computationally expensive. Conscience thresholds are architecturally immutable, they cannot erode through repeated exposure. The framework is resistant to all known jailbreak techniques and prompt injection attacks.
Your data stays yours. Core logic, memory, and audit logs run locally on your device. Choose cloud inference, local models like Gemma 4, or our zero-data-retention proxy. Three consent streams: Temporary, Partnered, or Anonymous, put you in control of what the agent learns and for how long per user (setup user is partnered by default). Full GDPR Article 15 and Article 17 support built in.
Works in your language. Native reasoning in 29+ languages including Amharic, Arabic, Hindi, and Swahili — genuine multilingual intelligence, not a translation layer.
$0.10/request with our hosted proxy, or free when you bring your own API key. Same price for everyone, everywhere.
Open source. Mission-locked. Built by CIRIS L3C, a low-profit LLC under AGPL-3.0. The entire stack — agent, proxy, billing — is open and auditable. 10 initial free interactions plus 2 free interactions per day are provided to support CIRIS L3Cs mission to demonstrate the viability of ethical AI.
30+ repositories. 1M+ lines of code.
https://ciris.ai
https://github.com/CIRISAI/CIRISAgent
CIRIS Agent is a personal AI assistant where every action passes through an 11-step ethical reasoning pipeline before execution. Four independent evaluations: common sense, domain-specific, principled, and intuition, plus a four-faculty conscience validate every decision. If the system isn't sure, it escalates to you. You always have final authority.
Every decision is cryptographically signed. CIRISVerify anchors your agent's identity in your device's Secure Enclave and signs actions with Ed25519 across five trust levels. The audit chain is hash-linked and tamper-evident. You can trace any reasoning chain and prove what happened.
Security that gets stronger over time. The Coherence Ratchet is our attempt to make honesty compound and deception computationally expensive. Conscience thresholds are architecturally immutable, they cannot erode through repeated exposure. The framework is resistant to all known jailbreak techniques and prompt injection attacks.
Your data stays yours. Core logic, memory, and audit logs run locally on your device. Choose cloud inference, local models like Gemma 4, or our zero-data-retention proxy. Three consent streams: Temporary, Partnered, or Anonymous, put you in control of what the agent learns and for how long per user (setup user is partnered by default). Full GDPR Article 15 and Article 17 support built in.
Works in your language. Native reasoning in 29+ languages including Amharic, Arabic, Hindi, and Swahili — genuine multilingual intelligence, not a translation layer.
$0.10/request with our hosted proxy, or free when you bring your own API key. Same price for everyone, everywhere.
Open source. Mission-locked. Built by CIRIS L3C, a low-profit LLC under AGPL-3.0. The entire stack — agent, proxy, billing — is open and auditable. 10 initial free interactions plus 2 free interactions per day are provided to support CIRIS L3Cs mission to demonstrate the viability of ethical AI.
30+ repositories. 1M+ lines of code.
https://ciris.ai
https://github.com/CIRISAI/CIRISAgent
Show More
