Privacy settings

Privacy and GDPR notice

Nfero uses essential storage for language preferences and demo access controls. With your consent, Google Analytics and lightweight usage events help improve model demos and outbound subscription routing. Live demo prompts may be sent to configured Alibaba-hosted model endpoints.

Agents and tools 8 min read

Individual Token Plan: Interactive Tools vs Automation

A dedicated API key does not mean unrestricted API use. Token Plan Individual is designed for one person's interactive agent workflow, with explicit boundaries around automation.

Who this guide is for

Developers deciding whether an interactive coding assistant or personal agent fits Token Plan Individual terms.

Who should use another route

Teams sharing credentials, SaaS backends, public chatbots, scheduled jobs, bulk generation pipelines or any workload that must run without a single user's active interactive session.

The key distinction is the workflow, not whether code makes the call

Token Plan Individual provides an API key, but Alibaba limits that key to personal, interactive use within programming and agent tools. The formal overview names tools such as Claude Code, Cursor, Qwen Code, Qoder and OpenClaw. The FAQ summarises the exclusion clearly: production automation, batch scripts and background scheduled tasks are not allowed.

An interactive tool can make several API calls for one user's task. Risk appears when the subscription serves other people, runs unattended, processes queues or powers an application's end users. Operating pattern matters more than programming language.

If the workflow must continue when the subscriber is absent, treat that as a warning sign and verify a production-appropriate billing route.

Examples that fit the documented interactive purpose

A solo developer can configure a compatible editor, terminal assistant or personal agent with the dedicated key and ask it to explain code, edit a repository, research a technical question or generate a media asset through an installed Skill. The person initiates the work, reviews the output and remains responsible for each session.

Alibaba's quick start lists Codex, Claude Code, Cursor, Qwen Code, OpenClaw and other clients. Compatibility still depends on protocol and model. Qwen3.8 and Qwen3.7 can use Harness tools through the Responses API; image and video models require a separate Skill or extension.

  • An individual asks a terminal agent to refactor a local project.
  • A creator invokes Wan image generation from a personal tool extension.
  • A user asks an agent to research a question with a supported Harness tool.
  • A developer reviews each request and output during an active session.

Examples that do not fit Token Plan Individual

Do not place the Individual key in a server route used by website visitors. Do not run a nightly script, monitor, queue consumer or bulk content generator with it. Do not share the key with colleagues or let several people use one account. Alibaba warns that prohibited use may lead to subscription suspension or key revocation.

An agent interface does not automatically make a workflow compliant. A public agent accepting customer jobs remains an application backend, while an hourly scheduler is background automation. Automation, team access or production availability requires a plan designed for that use.

  • Customer-facing chat, image or video generation.
  • Cron jobs, scheduled social posting or unattended monitoring.
  • Bulk evaluation, scraping or non-interactive batches.
  • A shared key in a team workspace.
  • Embedding the secret in a browser app, plugin bundle or public repository.

Keep the dedicated credentials and endpoint together

Individual keys begin with sk-sp-. Alibaba says the full key is shown only once, so use secure credential storage—never client-side JavaScript or committed configuration. The OpenAI-compatible Base URL is https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1.

Token Plan, Coding Plan and pay-as-you-go credentials are isolated. Mixing a key from one product with another product's Base URL can produce authentication or unsupported-model errors. Model names are also matched exactly against the Individual allowlist. Keep the key, endpoint, protocol and model ID as one reviewed configuration unit.

  • Use environment variables or the client's protected secret store.
  • Never log the full key or paste it into support screenshots.
  • Rotate or replace credentials if exposure is suspected.
  • Use separate configurations when comparing billing products.

Choose a billing route from the operating model

Individual fits one subscriber actively driving the work. Review Team Edition for several authorised users. Coding Plan has its own models, key and endpoint. Investigate pay as you go for production API access, automation or usage-based billing.

Do not assume the name of a plan grants a use right. Read the current terms and supported-model table for the specific product. A useful architecture review asks who initiates requests, who receives outputs, whether the system runs unattended, where the key lives and what happens when a quota pauses service.

Practical routing questions
RequirementRoute to review
One person's active agent sessionToken Plan Individual
Multiple people need authorised accessToken Plan Team Edition
Coding-focused subscription with its own allowlistCoding Plan
Backend, automation or customer-facing APIProduction pay-as-you-go terms

Who this is for—and who it is not for

Token Plan Individual is for a single person who wants predictable subscription access across supported text, image, video and tool calls while working interactively in compatible software. It can suit coding, research and creative exploration when the user remains in the loop.

It cannot fund a service's inference layer. Letting site visitors submit prompts through one Individual key would be backend use outside the documented terms. Use an approved production account, server-side secrets, rate limits, safety controls and suitable billing.

Frequently asked questions

Can an interactive agent call several models and tools in one task?

Yes, if the compatible personal tool supports those calls and the exact models and tools are included. Their usage draws from the same Credits limits. The session must still remain within the Individual plan's personal, interactive purpose.

Can I schedule OpenClaw to run while I am away?

The current FAQ prohibits background scheduled tasks and non-interactive automation. Using an agent product does not override that rule. Keep Individual use actively user-initiated, or choose a billing route whose terms allow unattended operation.

Can my website let users bring their own Individual key?

Do not assume that makes the workflow permitted. If the website operates as an application backend or automates requests, it can still fall outside Individual terms, and collecting user keys introduces serious security duties. Obtain written provider guidance before designing that feature.

Claims checked against official sources on Jul 23, 2026.

Official sources

Next step

Check which plan fits

Review current terms and availability before choosing a subscription or production route.

Continue on Nfero