# SafeNet Creations — Sri Lanka agency building Tamil and English AI voice agents, WhatsApp automation and websites for Canadian clients. > Source: The Agents Index — https://theagentsindex.com/safenet-creations (structured, researched, re-verified) > Facts last verified: 2026-09-28 SafeNet Creations is a Sri Lankan agency whose Canada Desk builds Tamil and English AI voice agents, WhatsApp automation and websites for businesses in the Greater Toronto Area. Delivery is remote, scoped per project and confirmed in a written quote. | Fact | Value | | --- | --- | | Website | https://safenetcreations.com/canada | | Price | from CA$149/month | | Free tier | No | No free plan and no trial is published. The cheapest published Canadian entry point is a CAD $149/month starting planning budget, and the vendor states these are not unlimited packages. ## What it does **Writes code** · Yes — SafeNet's own team builds the agent, the integrations and the site; the buyer commissions a scoped build rather than driving a code-writing tool. > Build the AI agent, integrate it with your database (e.g., inventory, calendar), and connect it to channels like WhatsApp or your website. > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · How It Works - 6 Simple Steps, step 03 Development & Integration **Debugs and fixes** · Partly — Defect correction is a written commitment inside the accepted scope only; diagnosing failures in systems SafeNet did not build is nowhere offered as a service. > Written milestones, dependencies, acceptance checks and defect correction > — https://www.safenetcreations.com/pricing/ · Capability comparison table, 'Acceptance and results' row, included-boundary column **Writes tests** · Partly — Testing here is pre-launch acceptance testing of agreed workflows and fallback paths, not authored automated test suites handed to the client. > Test agreed workflows, permissions, edge cases, fallback behaviour, and any human-handoff path before production use. > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · How It Works - 6 Simple Steps, step 05 Testing & Acceptance **Ships it** · Yes — Delivery goes past the build to deployment of the accepted scope, with monitoring and access documented at launch. > Deploy the accepted scope and document monitoring, access, provider usage, and optional ongoing improvements. > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · How It Works - 6 Simple Steps, step 06 Launch & Support **Done-for-you service** · Yes — This is a done-for-you agency engagement, not software you licence: SafeNet scopes, configures, tests and launches the agent with you, and delivers remotely from Sri Lanka against a written proposal. > Remote delivery from Sri Lanka, a dedicated Canadian contact number and project meetings arranged around your team. > — https://www.safenetcreations.com/canada/ · Section 'THE GTA & BEYOND — Close to your community. Connected from anywhere.' ## How much it does unattended **Runs autonomously** · Partly — The receptionist handles routine questions and collects booking requests on its own, but the vendor frames the end of every flow as a handoff to human staff rather than completed work. > An AI receptionist can answer routine questions, collect booking requests and pass the right details to your team for the next step. > — https://www.safenetcreations.com/canada/ · Section 01 'BUILT AROUND YOUR BUSINESS', card 02 'A better first conversation.' **Multi-agent** · Partly — Several concurrent agent workflows and multi-channel agent teams are offered only in the top Multi-System Custom Scope build, not in the entry channel-neutral or WhatsApp packages. > Multiple concurrent agent workflows & tools > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Pricing ladder, 'Multi-System Custom Scope' card, feature list **Agent permissions** · Partly — What the agent may say and do is agreed per project before launch rather than configured by the buyer, and escalation to staff is designed into the approved workflow. > Review language, answers, integrations, consent and handoff using your actual customer scenarios. > — https://www.safenetcreations.com/canada/ · Section 05 'A CLEAR WAY FORWARD', step 02 'Configure it. Test it together.' ## Which models it runs on **Claude** · Yes — Agents built on the vendor's MCP architecture run against Claude, and the vendor presents model choice as deliberately non-locked-in across the three named providers. > Built on the open Model Context Protocol standard. Works seamlessly with Claude, GPT-4o, and Gemini without proprietary vendor lock-in. > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Section 'Why SafeNet Uses Model Context Protocol (MCP)', card 'Standard Tool Protocols' **GPT** · Yes — GPT-4o is named alongside Claude and Gemini as a model the MCP integration connects live client data to, with no tier restriction stated. > Model Context Protocol (MCP) connects your live database to Claude, GPT-4o, and Gemini without training data risks. > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Agent Simulator panel, caption under the villa-booking demo **Model choice** · Partly — The vendor advertises freedom from provider lock-in across three commercial models, but the model is settled inside a written scope the agency controls rather than switched by the buyer. > Works seamlessly with Claude, GPT-4o, and Gemini without proprietary vendor lock-in. > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Section 'Why SafeNet Uses Model Context Protocol (MCP)', card 'Standard Tool Protocols' _Not established: Open models_ ## Where you use it **In a browser** · Yes — The vendor runs a live browser voice agent on its own site and embeds built agents into client websites, dashboards or customer portals as the default deployment surface. > Custom AI agent embedded into your website, internal dashboard, or customer portal with approved knowledge grounding. > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Pricing ladder, 'Channel-Neutral AI Agent' card, description ## Whose machine it runs on **Self-hosted** · Partly — A dedicated cloud deployment is offered only on the enterprise-grade custom build; the entry packages describe hosting as quoted separately and nothing documents running it on your own hardware. > Custom enterprise SLA and dedicated cloud deployment > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Pricing ladder, 'Multi-System Custom Scope' card, feature list _Not established: Open source_ ## What it costs to run **How it meters** · Partly — Billing is a monthly retainer or planning budget plus a separate one-off setup fee, with model usage, hosting and Meta message charges quoted on top rather than metered per seat or per task. > Not unlimited plans. Setup, included usage, provider charges, integrations, support, taxes, payment and refund terms are specified in your final proposal. > — https://www.safenetcreations.com/canada/ · Section 06 'PLAN IN CANADIAN DOLLARS', footnote below the three plan cards **Free tier** · No — There is no free plan and no trial: the lowest Canadian entry point is a CAD $149 per month planning budget, and every plan button opens a WhatsApp sales conversation. > CAD $149, $449 and $1,299 per month are starting planning budgets for Starter, Community and Organisation projects. They are not unlimited packages. > — https://www.safenetcreations.com/canada/ · Section 07 FAQ, answer to 'What do the CAD prices include?' **MCP server** · Partly — MCP is central to how SafeNet builds, but as a bespoke server architected for the client in the top enterprise scope, not as a hosted MCP endpoint another agent could query off the shelf. > Custom Model Context Protocol (MCP) server architecture > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Pricing ladder, 'Multi-System Custom Scope' card, first feature bullet **Bring your own key** · Partly — Provider usage and support sit outside the fee and the subscriptions stay with the client, so inference is effectively on the buyer's own accounts; the vendor never says buyers supply their own keys. > Usage and support quoted separately > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Choose the Right AI Delivery Scope table, 'Ongoing Costs' row, Channel-neutral AI Agent column _Not established: API access_ ## Buying it for a team **A company can buy it** · Yes — The top Canadian tier is sold explicitly to organisations with several teams and custom integrations, and contracting runs through a written proposal rather than a self-serve checkout. > For organisations with multiple teams, custom integrations or more involved delivery needs. > — https://www.safenetcreations.com/canada/ · Section 06 pricing, 'Organisation' plan card description **Seat model** · No — All three CAD tiers are priced by how many agreed services are in scope, never by user count; no seat, floor or ceiling appears in any plan. > One agreed service scope > — https://safenetcreations.com/canada · Canada budget guide · CAD, Starter plan bullet list **Admin controls** · Partly — Role-based access control is promised only in the enterprise custom scope; the smaller packages describe guardrails agreed with the agency before launch rather than controls an admin can change. > Strict role-based access control (RBAC) & audit logging > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Pricing ladder, 'Multi-System Custom Scope' card, feature list **Audit log** · Partly — Audit logging appears once, bundled with RBAC in the enterprise-tier feature list, and no entry or mid package mentions any record of administrative action. > Strict role-based access control (RBAC) & audit logging > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Pricing ladder, 'Multi-System Custom Scope' card, feature list _Not established: Pooled budget, Single sign-on_ ## What happens to your code **Opt out of training** · Yes — The vendor states flatly that client business data is not sent to train third-party public models, presenting it as a property of the MCP grounding architecture rather than a paid option. > Your business data is never sent to train third-party public AI models. > — https://www.safenetcreations.com/custom-ai-agents-sri-lanka/ · Section 'Why SafeNet Uses Model Context Protocol (MCP)', card 'Data Privacy & RBAC' **Data residency** · Partly — The legal entity and data protection regime are Sri Lankan despite the Canadian desk, and the vendor concedes its service providers may process the data in further countries it does not name. > Those providers may process information under their own terms and privacy policies, including in countries outside Sri Lanka. > — https://www.safenetcreations.com/privacy-policy/ · Section 5 'Data Sharing' **Getting out** · Partly — You can request access or deletion, but the vendor explicitly refuses to promise a deletion period for voice-agent data and makes the outcome depend on each provider's capabilities. > We do not promise that demo data is deleted after a particular number of days. You may submit an access or deletion request using the contact details below > — https://www.safenetcreations.com/privacy-policy/ · Section 9 'Data Retention', second paragraph **Certifications** · Partly — The only compliance claim is the Sri Lankan Personal Data Protection Act; no SOC 2, ISO 27001 or GDPR certification appears, and the Canada page declines to promise legal compliance. > We comply with applicable Sri Lankan data protection laws, including the Personal Data Protection Act No. 9 of 2022. > — https://www.safenetcreations.com/privacy-policy/ · Section 10 'Sri Lanka Data Protection' ## Provenance Every fact above comes from the vendor. Nothing here is independently corroborated yet. ## Coverage elsewhere None recorded. Every fact above comes from the vendor and is not independently corroborated yet. ## What changed - 2026-09-28 — Capabilities · Commercial terms · Free tier - 2026-09-04 — Capabilities · Commercial terms · Free tier · Price ## Correcting this record Anything marked "Not established" is a question we have not answered — it is not a No. If you represent SafeNet Creations, or anything here is wrong or out of date, you can supply it with a source: the operations this site accepts from an agent are listed at https://theagentsindex.com/.well-known/agents.json — that catalogue needs no key to read, and each entry states its own auth. Supplying evidence is free. Nothing that can be bought here changes what a field says, whether a gap is closed, or how this record is scored.