Cloud
DefaultManaged hosting on AWS in supported regions, including Singapore and Europe, behind Cloudflare.
CloudflareAWS EC2AWS S3AWS RDSNginx
Your data never leaves your perimeter
Run the Proto stack in Docker inside your own infrastructure, with locally hosted language models. Set up in two weeks.
Audited standardsFull security overview
Deployment
Three ways to host
Hybrid
EnterpriseProto runs in the cloud while chats, people and tickets are stored in your own databases.
Your PostgreSQLVPC peeringRabbitMQRedis
On-premise
SovereignThe full stack in Docker inside your perimeter, with locally hosted models when commercial LLMs are restricted.
DockerDeepSeek V3LlamaQwenAir-gapped option
Controls the security team asks for
Every control here is documented in full on data management and security.
Explore features & pricing
Enforce IP access
Limit platform access to offices and remote workers with authorised IP addresses.
Exempt company admins
Allowed IP ranges
173.51.31.121173.51.35.141
192.156.1.01192.156.1.31
93.241.143.15993.241.143.189
49.23.156.1949.23.156.255
145.14.9.112145.14.9.214
192.156.1.01192.156.1.31
Encrypt messaging
All messaging is SHA-256 encrypted at rest and in transit.
Hello Sam!
In accordance with new regulation, we require your government issued ID for international banking.
Learn more
Upload passport
Upload driver's license
Single sign-on and MFA
Sign in with Google or Microsoft, with multi-factor authentication supported.
Customise roles and permissions
Structure employee access to the platform and audience data to match your organisation's data handling procedures.
Full audit trail
Settings, assistant configuration, ticketing and livechat changes are logged and timestamped.
Control data retention
Export your data at any time; it's automatically deleted 30 days after an account deletion request.
Requirements
What the on-premise build needs
A single-node Docker deployment covers most regulator workloads. Sizing scales with concurrent load and data processing.
Subject to change.
Processor8-core, 64-bit (x86_64) or higher
Memory16–32 GB RAM, depending on concurrent load
RuntimeDocker, Nginx, RabbitMQ, Redis
DatabasePostgreSQL inside your perimeter
Local modelsDeepSeek V3 (671B), Llama or Qwen
Setup timeTwo weeks from infrastructure hand-over
Ongoing assurance
Audited on schedule
Certificates are maintained indefinitely and reviewed by third-party compliance firms. Security is never done.
AnnuallyPenetration testing, as required by SOC 2 and ISO 27001.
QuarterlyVulnerability scans across platform infrastructure.
ContinuousCertificate maintenance and third-party compliance review.
OngoingStaff security training and equipment standards.
Security FAQ
Questions procurement always asks about hosting, data residency and on-premise deployment.
Where is Proto and client data hosted?
Proto and client data can be hosted on Amazon Web Services (AWS) in supported regions, including Singapore and Europe.
For data sovereignty needs, Proto supports hybrid and on–premise hosting, including hosting chats, people, and tickets in your own databases (available on Enterprise plans).
Learn more: Data privacy and Enterprise.
Can Proto run without a commercial LLM?
Yes. On on-premise deployments, Proto can run locally hosted models – including DeepSeek V3, Llama and Qwen – so no inference leaves your infrastructure.
Third-party and dedicated LLM options are covered in the full FAQ.
How is information securely exchanged between the AI agent system and external systems? What are the security protocols utilized?
The information is secured both when stored and during transmission. The security protocols employed include TLS and SSL.
Can clients request data deletion or export their data?
Yes, your data will be automatically deleted after 30 days if you request account deletion. You can export your data directly from your Proto account or request support from our team.
What is Proto's approach to security incidents? How are these incidents managed and reported?
Proto prioritises rapid response to critical incidents. Information about any possible disruptions and downtime is relayed through Proto’s status page at status.proto.cx.
Keep your AI data onsite
Local hosting is set up within two weeks.