Maple AI
OpenSecretEnd-to-end encrypted AI chat assistant. Conversations are encrypted on the device and processed inside hardware-isolated Trusted Execution Environments, on AWS Nitro Enclaves with Nvidia GPU TEEs, so the operator cannot read them; remote attestation lets users check the running build against published source. It serves open-weights models and ships as web, iOS, Android and desktop apps, built on the OpenSecret confidential-computing platform.
The consumer service at trymaple.ai is paid; openness here describes the code and architecture rather than free usage. Verified 2026-08-13 via Maple's model guide and its first-year retrospective.
Openness
5 medium confidence- license
- MIT(Maple-client)+AGPL-3.0(OpenSecret-platform),both-OSI
- source
- public(github.com/OpenSecretCloud)
- core-gated
- ungated(billing and feature-flag APIs are optional and external
- models
- open-weight
- hosted-tier
- paid(managed-enclaves)
Both halves of the stack are open under standard OSI licenses - the Maple client under MIT, the OpenSecret platform under AGPL-3.0 - and the served models are open-weight, which is unusually clean for a "private AI" product. It scores as open source because the privacy guarantee is only credible where the code is open and attestable. The paid tier is not a withheld core: the Maple README makes the billing and feature-flag services optional external clients, configured through VITE_MAPLE_BILLING_API_URL and VITE_OS_FLAGS_BASE_URL, and the OpenSecret README says those "server implementations are not part of this repository setup". A self-hosted backend runs without either, so what the subscription buys is usage on the hosted enclaves rather than functionality missing from the source, and neither repository has an enterprise directory or a license key. The one closed-looking artifact is a 15MB continuum-proxy binary vendored into OpenSecret, and that is the third-party trusted-execution proxy from Edgeless Systems, carried alongside a submodule of edgelesssys/privatemode-public rather than anything OpenSecret withholds from a self-hoster.
- https://github.com/OpenSecretCloud/Maple recorded 2026-06-17
MIT-licensed Maple client (Tauri/Rust/TS), active; multi-platform
- https://github.com/OpenSecretCloud/opensecret recorded 2026-06-17
AGPL-3.0 OpenSecret platform; AWS Nitro enclave / TEE confidential-computing architecture
- https://github.com/OpenSecretCloud/Maple/blob/master/LICENSE recorded 2026-08-12
the Maple client repository LICENSE is a plain MIT License; the GitHub repo API reports the SPDX id MIT
- https://github.com/OpenSecretCloud/opensecret/blob/master/LICENSE recorded 2026-08-12
the OpenSecret platform repository LICENSE is the GNU Affero General Public License Version 3; the GitHub repo API reports the SPDX id AGPL-3.0
- https://github.com/OpenSecretCloud/Maple/blob/master/README.md recorded 2026-08-12
the configuration surface lists VITE_OS_FLAGS_BASE_URL as "an optional feature-flags API" and VITE_MAPLE_BILLING_API_URL as "an optional billing API", and states "Flags and billing are independent clients"; the quick start runs the client against a local OpenSecret backend at 127.0.0.1:3000 with no paid component
- https://github.com/OpenSecretCloud/opensecret/blob/master/README.md recorded 2026-08-12
the backend is "the open-source Rust backend for confidential AI applications such as Maple" and builds from the Nix flake; "Billing and feature flags are optional external HTTP APIs whose administrative credentials remain in the backend; their server implementations are not part of this repository setup"
Adoption
3 low confidenceLaunched Jan 2025; vendor reports "5x user growth in the last six months", >90% month-over-month paid retention, and "billions of tokens" in year one, but no absolute user count is published. Self-reported figures only, hence low confidence.
- https://blog.trymaple.ai/one-year-of-private-ai-how-maple-ai-made-encryption-easy/ recorded 2026-08-13
launch date (Jan 2025); 5x growth, >90% retention, billions of tokens processed
Capability
4 medium confidenceBroad multi-model, multi-platform private chat with vision, API, and teams. Short of top because it is a focused private-chat product without a confirmed agentic tool-use/web-browsing ecosystem.
- https://blog.trymaple.ai/maple-ai-model-guide-with-example-prompts/ recorded 2026-08-13
model lineup (open-weight) and tier/feature structure
Verified 2026-08-12