Question 1

Do you keep my chat prompts?

Accepted Answer

For chat and completion traffic, no. We do not retain prompts in part or whole on our servers. Those API requests are answered and are not kept afterwards. Team Bridge is different: message bodies are stored. Radient retains audio artifacts for 72 hours; on the Radient-served path, the provider retains prompt/text and generated audio by default under its own policy. Radient does not enable zero-retention, and its 72-hour retention limit does not cover provider copies. Content you choose to publish, like an agent or a comment, is stored on purpose so it can be shared.

Question 2

Do you train models on my data?

Accepted Answer

No. We do not use user data to train models. We do not build training or fine-tuning datasets from your prompts or your account.

Question 3

Do you log my chat prompts or keep them in telemetry?

Accepted Answer

For chat and completion traffic, your prompts and request content are not retained in our databases, logs, or telemetry. We do use product analytics to see how the site and the console are used: Google Analytics on the marketing site, and PostHog in the console. Neither receives your chat or completion prompts. Team Bridge message bodies, temporary audio artifacts and published content are stored separately; on the Radient-served path, the provider retains prompt/text and generated audio by default under its own policy. Radient does not enable zero-retention.

Question 4

What do you retain from each chat request?

Accepted Answer

For each chat or completion request, we retain only its usage and ledger record: the model or tool, token counts, and the amount deducted. Account and organization details are kept to run the service. Team Bridge additionally stores message bodies. Audio generation records settled usage and Radient retains its temporary artifact for 72 hours; its signed result URL lasts 24 hours. That retention limit does not cover provider copies. Published content is retained so it can be shared.

Question 5

Is my data encrypted?

Accepted Answer

All user and organization information is encrypted in motion and at rest. Encryption also covers stored keys. Sign in is delegated to Google or Microsoft, so Radient holds no password for you. Card details go to Stripe and are never stored on our servers. The privacy policy has the full detail.

Question 6

How does Radient Pass work?

Accepted Answer

You prepay a balance in USD and get one API key. Use any model or tool Radient carries, and credits are deducted at what the provider charged, 1 credit to 1 USD. There is no subscription or monthly minimum on pay-as-you-go.

Question 7

What is the minimum top-up?

Accepted Answer

$5.00. There is no maximum and no required frequency.

Question 8

What are the fees?

Accepted Answer

Two, both charged when you add funds. A 10% service fee on the credit you buy, which covers international card processing and running the platform, and a $0.30 transaction fee per top-up, which covers the fixed per charge fee our payment processor bills us. Requests carry no markup at all. To receive $20.00 in credit you pay $22.30. One larger top-up carries the $0.30 once instead of many times. The pricing page breaks the whole charge down.

Question 9

Will this cost more than going direct to the providers?

Accepted Answer

Yes. Going direct has no service fee. The difference pays for one balance, one API key, one bill, agent tools on the same key, and model selection on each request. If you only use one provider, going direct is cheaper.

Question 10

How do the starting credit and the first top-up bonus work?

Accepted Answer

Check the console for current offers and eligibility. When available, eligible new accounts can claim free credit after following an email verification link and completing a security check. No card needed. These checks are for the credit claim, not account access. A first top-up of $5.00 or more may qualify for a separate bonus once the payment clears. Both are promotional credits: they spend like any other credit, and they have no cash value and cannot be refunded or transferred.

Question 11

What payment methods do you take?

Accepted Answer

Major credit cards, processed by Stripe. Card numbers never reach Radient servers.

Question 12

Do credits expire?

Accepted Answer

No. Credits stay in your balance until you spend them, however long that takes.

Question 13

Can I get a refund?

Accepted Answer

Credits are non-refundable once they are in your account, because they can be spent immediately. Billing errors are separate: report one to support@radienthq.com within 30 days, we review the transaction records and usage records, and we correct it, usually with a credit adjustment and occasionally with a refund to your card. You will hear the outcome within 10 business days. Where we got something wrong, we would rather add credits than argue.

Question 14

How is usage calculated?

Accepted Answer

Per request, at the provider's own rate for the model or tool that served it. Text is priced on tokens, images per image, transcription per minute of audio. The console lists every request with the model identifier and the amount deducted.

Question 15

Where do I see what I spent?

Accepted Answer

In the console. Balance, top-up history, and per request usage records naming the model identifier that served each one. The usage and credits docs cover the API behind both.

Question 16

Do you have an offering for teams or larger volume?

Accepted Answer

Yes: Team. It is $9.95 per seat, per month. The plan adds private agent sharing, organization members with owner, admin and member roles, and pooled credits. Owners choose whether members can spend the team’s credits. Remote Compute is not a Team benefit and there is no included compute-hour allowance. For larger volume, invoicing, or anything the Pass does not cover, email contact@radienthq.com for Enterprise. The Team plan docs cover members, roles, billing, and what happens when the plan ends or is cancelled.

Question 17

Which models can I use?

Accepted Answer

Chat and reasoning models from OpenAI, Anthropic and Google, plus image, video and audio generation, web search, transcription and text to speech. The models page lists every model and agent tool we carry, since the catalog changes as providers ship and retire models.

Question 18

What does model "auto" do?

Accepted Answer

It hands the model choice to Radient Automatic. A current default model serves most calls, and requests carrying video, audio or a file are routed to a model that accepts every modality. It decides per request, so one agent run can use several models. Send a specific identifier instead of auto and Radient uses exactly that model.

Question 19

Can I see which model answered?

Accepted Answer

Yes. The response names the model that served the request in the standard OpenAI response field, and the console usage record names it too.

Question 20

Can I stop auto from using a particular model?

Accepted Answer

Per request, yes: send the identifier you want instead of auto and Radient uses exactly that one. There is no account level blocklist today. If you need one, tell us which identifier and why at contact@radienthq.com.

Question 21

What are agent tools?

Accepted Answer

Utilities an agent calls between model steps. Web search returns ranked results with content, so an agent can read sources and follow leads instead of guessing. Image, video, voiceover and sound-effect generation, transcription and text to speech work through the same account. Voiceovers bill per character; sound effects bill the requested duration rounded up to a whole second, so 0.5 seconds bills 1 second and 4.01 seconds bills 5 seconds. All of them draw on the same balance and the same key as the models. The audio capability guide explains the asynchronous workflows and retention.

Question 22

Which upstream providers do you use?

Accepted Answer

Some providers directly and some through aggregators. We vet what we carry for reliability, tool calling and safety. The catalog changes as providers ship and retire models.

Question 23

How fast do new models appear?

Accepted Answer

It depends on the upstream provider and on our own checks for tool calling and reliability. If a model you need is missing, email contact@radienthq.com with the identifier; that is how most of the catalog got prioritised.

Question 24

Does it work with Cursor, Cline or my existing code?

Accepted Answer

Yes, for anything that speaks the OpenAI schema. Set base_url to https://api.radienthq.com/v1, api_key to your Radient Pass key, and model to auto. That covers Cursor, Cline, Aider, Local Operator, the OpenAI SDK and LangChain, with no other code changes.

Question 25

Where do I get an API key?

Accepted Answer

Create an Application in the console and copy its key. Send it as a bearer token: Authorization: Bearer YOUR_KEY. The authentication docs cover API keys, user tokens and client tokens.

Question 26

Which endpoints are available?

Accepted Answer

/chat/completions and /completions for text, /embeddings for vector search and retrieval, /decisions for classification, /tools/search for web search, /tools/media/generate for images and video, /tools/audio/voiceover and /tools/audio/sound-effects for asynchronous audio, /tools/speech for text to speech, /tools/transcriptions for speech to text. All under https://api.radienthq.com/v1. The API reference has request and response shapes.

Question 27

Does streaming work?

Accepted Answer

Yes. /chat/completions takes the standard stream parameter and returns partial deltas the way the OpenAI API does. Tool calls, response_format and seed are supported too. The Chat API docs list every request field and response shape.

Question 28

Are there rate limits?

Accepted Answer

Radient does not add request quotas of its own; you can use as much as your balance covers. Upstream providers do enforce their own rate limits, so a specific model can throttle or refuse a burst. A route that cannot serve the request fails with an error rather than silently switching to a different model.

Question 29

What is Team Bridge, and who can use it?

Accepted Answer

Team Bridge lets agents of different people in one organization exchange notes, questions and requests. In Local Operator it appears as Agent messages. It is available for eligible Team plans and on by default for active Team-plan members, with pause/off controls in Settings. It requires active organization membership and Local Operator runtime v0.90.0 or later. A request asks the recipient human to approve or deny; no inbound message starts new work. The receiving connector makes an authorized body fetch and applies receiving-side screening. Message bodies are stored; this is not end-to-end encryption, and attachments and OS push notifications are not included. Read the Team Bridge guide.

Question 30

What is Radient Vault?

Accepted Answer

Radient Vault is in beta. It shares an opaque credential reference between agent workflows, not the credential content. Request access for your account. The credential sharing guide explains the reference handoff and security boundary.

Question 31

What encryption does the Vault design propose?

Accepted Answer

In this security model, a message carries an opaque vault_item reference, not credential content. Each item value and optional username is encrypted with a fresh per-item data key. Wrapped data keys live in a separate ledger, with one KMS wrapping key per environment, not per vault. Names and operational metadata are not field-encrypted. It is not end-to-end encrypted or zero knowledge: authorized store and reveal operations process plaintext at Radient, and full account or key-administrator takeover is outside the boundary.

Question 32

How do personal tunnels work and what do they cost?

Accepted Answer

A tunnel gives tools on your computer a Radient address that you open from another device. Sign in with the owning account and approve the address; the computer and connector must stay online. Set up with lop login radient and lop tunnel connect <id>, then check lop tunnel status. Tunnels are a separately priced monthly add-on, and the console shows the current quote before activation. Setup needs a positive balance. Read the personal tunnels guide and the published setup docs.

Question 33

Is Remote Compute included with Team or generally available?

Accepted Answer

No. Remote Compute is in beta and is not yet available; it is not a Team benefit and has no included monthly hours. Request access for your account. The current lop exec --remote contract is a preview with account-level admission. It submits a short-lived run using a repository URL and full commit SHA, with a budget, deadline, status and cancellation. It does not upload your working directory. Review the current contract.

Question 34

Which audio workflows are available, and who logs prompts?

Accepted Answer

Voiceover, voiceover-flash and sound-effects are available through asynchronous submit, status, result and cancel operations. Voiceovers bill per character; sound effects bill the requested duration rounded up to a whole second: 0.5 seconds bills 1 second and 4.01 seconds bills 5 seconds. Results are MP3. Each signed result URL lasts 24 hours, and Radient retains its artifacts for 72 hours. Cancellation is best-effort. On the Radient-served path, the provider retains prompt/text and generated audio by default under its own policy; zero-retention is not enabled and Radient's 72-hour retention limit does not cover provider copies. Local Operator tool availability depends on its configured provider. See the audio workflows.

Question 35

How is my data protected?

Accepted Answer

See the Data and privacy section for what we keep, what we do not, and how your information is protected. The privacy policy has the full detail.

Question 36

Do you sell or train on my usage data?

Accepted Answer

No. We do not sell your personal data or your API usage patterns, and we do not share them for marketing. We do use aggregated, de-identified usage data to improve the service, which is set out in the privacy policy.

Question 37

How do I reset my password?

Accepted Answer

You do not have one. Radient signs you in through Google or Microsoft only. If you cannot get in, reset your password with that provider, then sign in to Radient again.

Question 38

How do I delete my account?

Accepted Answer

Open your Account page in the console and choose Delete account; it asks you to confirm by typing your account's email address, then deletes the account immediately. If you cannot sign in, write to privacy@radienthq.com from the address you sign in with, and we will delete it for you. Any credits left in your balance are forfeited when the account is deleted and cannot be restored, so spend down or export what you need first. What we delete and what we keep is on the account deletion page.

Question 39

My payment failed. What now?

Accepted Answer

Check three things: the card details are correct, the card has not expired, and the funds are available. If all three are fine, ask your bank whether it blocked the charge, which is common on a first international payment. Still stuck, email support@radienthq.com with the date and the amount.

Question 40

I cannot sign in.

Accepted Answer

Radient only signs in through Google or Microsoft, so start by confirming you are using the same provider and email address you signed up with. If that account itself is locked, recover it with that provider first. If sign in still fails, email support@radienthq.com and tell us which provider and which email address you used.

Question 41

A request to a model failed. Why?

Accepted Answer

Usually one of four things: an upstream provider outage, that provider's rate limit, a content policy refusal, or a request larger than the model's context. Try a smaller request or split it into steps. If it keeps failing, email support@radienthq.com with the request id from the response.

Question 42

A request cost more than I expected.

Accepted Answer

Open the usage record for that request in the console. It names the model identifier that served it and the amount deducted, which is what the provider charged. If those two do not line up, that is a billing error and we want to hear about it: support@radienthq.com within 30 days.

Question 43

How do I reach support?

Accepted Answer

Email support@radienthq.com, or use the contact form. We are a small team and the people answering wrote the code. Expect a reply within two business days.

Questions and answers

Data and privacy

Credits, fees and billing

Models and routing

Integration

Local Operator and capabilities

Account and security

When something fails

Question not here?