Ptero

Ptero (ptero.pro) is a browser-based AI chat interface: pick a display name, pick a model, start chatting, no account or payment required. It is built and shipped as a single WordPress plugin (the live site is literally WordPress under the hood), and as of this check it offers nine chat models through a dropdown picker: Ling 3.0 Flash, Deepseek v4.1 Flash, Atria-Dawn-Preview, Agnes 2.5 Flash, Agnes 3.0 Flash, Claude Haiku 5.5, Mistral Nemo 12B, Laguna S 2.1 and Apodex 1.1 Mini (source: live model picker and in-app API docs on ptero.pro, 2026-10-10, https://ptero.pro/). The GitHub repository behind it (Ptero-ai/Ptero) was created 2026-08-18 and last had code pushed 2026-10-07, so the project is about seven weeks old at time of writing (https://api.github.com/repos/Ptero-ai/Ptero).

Ptero's own Terms of Service state: "Ptero.pro is completely free to use and will always be free. We do not offer, and will not introduce, premium plans, subscriptions, paywalls, or any paid tier. Every model made available through the Service is provided to you at no cost. We reserve the right to change which specific models are offered, but not to charge for access to the Service itself." (https://ptero.pro/, Terms of Service modal, accessed 2026-10-10). That is a stronger and more specific promise than most "free AI chat" sites make — most reserve the right to add paid tiers later. The GitHub README separately describes Ptero as a "non-profit project"; that framing is the vendor's own and we found no registered non-profit entity to corroborate it, so treat it as a claim rather than a verified status.

Per Ptero's Privacy Policy (last updated 2026-08-15): chat content passes through their server only momentarily to relay the request to the chosen AI provider and "is not written to our database"; conversation history lives in your browser's local storage, and clearing browser data or switching devices removes it (https://ptero.pro/, Privacy Policy modal). That is a genuinely unusual design for a free hosted chat tool — most log everything server-side by default. The carve-out: the in-app onboarding and consent screens separately disclose that chats inside a "Project" (Ptero's workspace/grouping feature) are synced to their servers so they're available across devices, with a note to "export a backup first" before clearing local data (https://ptero.pro/, first-run and consent modals). The formal Privacy Policy text doesn't call out that Project carve-out explicitly, so a user relying only on the policy page could miss it.

Some models on the list are easy to place: Claude Haiku 5.5 is Anthropic's small model released 2026-10-07 (https://www.anthropic.com/claude-haiku-5-5), meaning Ptero added it within about three days of release. Deepseek v4.1 Flash and Mistral Nemo 12B are likewise traceable to DeepSeek and Mistral respectively, and Atria-Dawn-Preview traces to a small research outfit (atominnolab.com's "Atria" project, https://www.atria-asi.ai/). For three others on the roster — Agnes 2.5 Flash, Agnes 3.0 Flash, Laguna S 2.1, and Apodex 1.1 Mini — we could not find any AI lab, model card, or announcement matching those names anywhere online. That doesn't mean they're fake; it means their provenance is not established, and a buyer who cares about knowing which company actually trained the model they're talking to should stick to the identifiable names on the list.

Apodex 1.1 Mini is marked "Star to unlock" in the model picker: using it requires signing in with GitHub and starring the Ptero-ai/Ptero repository, done through an in-app OAuth flow that adds the star for you (https://ptero.pro/, model picker and star-gate modal). This is a growth mechanic rather than a cost, but it does mean one of the nine models requires linking a GitHub account to use, which is worth knowing before you pick a model expecting the same no-account experience as the rest of the site.

The code is on GitHub at Ptero-ai/Ptero: 1,496 stars, 15 forks, 0 open issues, last pushed 2026-10-07 (https://api.github.com/repos/Ptero-ai/Ptero). It is not archived. GitHub's own license detection, reading the actual LICENSE file in the repo, identifies it as the Apache License 2.0 — but the README's badge and license section both describe the project as "GPL v2 or later." We checked the LICENSE file directly and it is the standard Apache-2.0 text (https://github.com/Ptero-ai/Ptero/blob/main/LICENSE); anyone planning to fork or redistribute should go by that file, not the README's description of it. The repo itself is two PHP files — a short plugin bootstrap and a single ~1.5MB file holding the bulk of the application — consistent with a monolithic WordPress plugin rather than a thin or incomplete mirror of the live site.

The project has no named company. Its GitHub organisation (Ptero-ai) and its sole visible contributor account (aminkheddache-dotcom, display name "Amine KHD") was created in December 2025 (https://api.github.com/users/aminkheddache-dotcom). Ptero's Terms of Service route contact to "the Service operator through pterocos.eu.org" — that address is a free .eu.org subdomain hosting a different, unrelated side project by the same operator (an online browser code editor, "Pterocos"), not a company site (https://pterocos.eu.org/). The GitHub README credits two sponsors, Agiler (a serverless WordPress hosting provider) and JTTI (a VPS/hosting reseller), for "keeping Ptero free"; Agiler's own site confirms it is a WordPress hosting company, and the author's schema.org profile links to an Agiler-hosted subdomain, consistent with Ptero running on Agiler's infrastructure (https://www.agiler.io/). This is a young, single-developer project without a registered business behind it — a continuity consideration like any early-stage free tool, not a red flag on its own.

Ptero exposes an OpenAI-compatible REST endpoint at https://ptero.pro/wp-json/mlp/v1/chat, with bearer-token authentication; API keys require GitHub verification to create (source: in-app API documentation panel, ptero.pro settings, accessed 2026-10-10). The worked cURL/JavaScript/Python examples in that panel are specific to Ptero's own endpoint and model IDs. A later block in the same panel, headed "Support & Resources," still shows unedited template placeholders — Documentation: https://docs.example.com, Status Page: https://status.example.com, GitHub: https://github.com/example/api-sdk, Discord: https://discord.gg/example — and a "Rate Limits & Quotas" paragraph (40,000–100,000 tokens/hour, varying by model) sits in that same unedited block, so we are not treating those specific numbers as confirmed current limits.