AI agents are starting to do real work inside education products: drafting lessons, checking answers, preparing training material, and turning raw prompts into structured learning content. But most education platforms still assume someone is sitting in front of a dashboard.
That creates a gap. A teacher may want a visual editor. A developer may want a REST API. An AI agent or CI workflow needs something else entirely: a command-line interface that can run without a browser, return structured output, and fail in a way another system can understand.
TutorFlow CLI was built for that layer.
It gives developers, AI agents, and automation scripts a direct way to use the TutorFlow Content API from the terminal. Instead of wiring every endpoint by hand, an agent can create slides, videos, courses, modules, tests, games, and AI evaluations through a stable command-line contract. The full command surface is documented in the TutorFlow CLI overview.
Why a CLI Matters for AI-Powered Education Workflows
The next generation of learning software will not be operated only by people clicking through screens. It will also be operated by agents that plan, generate, check, retry, and hand work back to humans when needed.
That changes what infrastructure needs to provide. An agent-friendly education API cannot rely on hidden UI state, one-off browser sessions, or terminal output that has to be scraped. It needs predictable inputs, structured output, and clear error behavior.
TutorFlow CLI turns those needs into a practical workflow:
- use
TUTORFLOW_API_KEYwith atf_content_key for environment-based authentication; - use
--jsonorTUTORFLOW_JSON=1for machine-readable output; - pass required inputs as flags or positional arguments;
- receive JSON errors on
stderr; - rely on stable exit codes for success, invalid input, cancellation, and failures;
- add
--waitwhen the next step needs the finished course, deck, or test rather than a run id.
For developers, that means less API boilerplate. For agents, it means fewer ambiguous states. For education teams, it means content generation and assessment workflows can move from manual experiments into repeatable operations.
What TutorFlow CLI Can Create
TutorFlow CLI covers the content and assessment workflows of the TutorFlow Content API. Everything it creates lands in a TutorFlow classroom, where a teacher can open it in the editor or share its public page.
It can generate multi-lesson courses from a prompt, slide decks from a topic or lesson brief, and interactive modules for reusable lessons. It can build narrated educational videos from the scenes you script and render them to mp4, and it can plan and build games and simulations. The supported commands are covered in the CLI docs.
It can also generate tests and assessments, then run AI evaluation and grading for exact-match, rubric-based, and open-ended learner responses. Evaluations tagged with concepts feed a learner model, so an agent can ask which concepts a learner has mastered. That combination is useful for product teams building tutoring, practice, assessment, onboarding, or internal training workflows.
Here is the shape of a typical CLI session:
# One-time: create an agent account; the tf_content_ key is saved locally.
tutorflow agent register --name "Ops Academy Agent" --slug ops-academy-agent --email ops@example.com
tutorflow --json course generate "Beginner Python for operations analysts" \
--lessons 5 --language en --wait
tutorflow --json evaluate \
--type open_ended \
--question "What causes seasons on Earth?" \
--student-answer "Earth's tilt changes how sunlight reaches each hemisphere." \
--max-score 10 \
--learner-id analyst-42 \
--concept-tags axial-tilt,seasons
tutorflow --json evaluation mastery analyst-42Generation runs in the background on TutorFlow's side. Without --wait, course generate returns a run id at once, and tutorflow run wait <runId> picks it up later, which suits agents that start several jobs and collect them afterwards. Every result carries an editorUrl for the classroom editor and, once published, a publicUrl for learners; tutorflow <type> share <id> makes a result open by link and prints that publicUrl.
The goal is not to replace the full TutorFlow web experience. The goal is to make the same education infrastructure callable from wherever work is happening: a local terminal, an AI coding agent, a CI job, or an internal operations script.
For agents that need to inspect platform status or prices before creating content, the CLI platform discovery docs explain the manifest, pricing, and health commands, and tutorflow credits balance shows what is left to spend.
Built for Non-Interactive Work
Many command-line tools are easy for humans but awkward for automation. They prompt unexpectedly, mix logs with data, or fail with messages that are readable to people but hard for software to classify.
TutorFlow CLI takes the opposite approach for agent workflows.
When JSON mode is enabled, success output goes to stdout as structured JSON. Errors go to stderr as JSON, with the API's own error code and a hint where one helps: an account out of credits, for example, gets content_payment_required and the command that opens a checkout. Missing required input in a non-interactive environment returns a clear failure instead of waiting forever for a prompt.
That matters when TutorFlow is part of a larger pipeline. A script can generate a module, store its returned ID, pass that ID into another system, and stop cleanly if the API key is missing, the request is invalid, or the generation run fails.
For commands that may be retried, idempotency keys help protect against duplicate asset creation:
tutorflow --json module generate "Workplace cybersecurity basics for new employees" \
--language en \
--idempotency-key onboarding-cybersecurity-v1 \
--waitThis is the detail that makes a CLI useful in production. The impressive part is not that it can create content once. The important part is that it can be trusted inside a repeatable workflow.
Where It Fits in an EdTech Stack
TutorFlow CLI is especially useful for teams that already have content operations, internal training processes, or product workflows that need to become more automated.
An EdTech product team can prototype AI course generation before committing to a deeper backend integration. A training team can generate draft modules from onboarding topics and route the editor links to human reviewers. A tutoring product can run structured answer evaluation as part of a test harness and read back each learner's mastery. An AI agent can check its credit balance, generate learning assets, render a narrated video, and return public links to an operator.
That makes TutorFlow CLI both a developer tool and an automation bridge. It sits between the API and the people, agents, and systems that need to use it quickly.
If an agent needs to register an account, check billing status, or create a hosted checkout session, those workflows are documented in CLI agent account and billing.
Why This Matters Now
AI-generated learning content is moving from novelty to infrastructure. The question is no longer whether AI can draft a quiz or outline a course. The harder question is whether those capabilities can be used reliably inside real systems.
Reliability comes from contracts: stable commands, structured output, explicit inputs, predictable errors, and authentication that does not depend on a person being present.
TutorFlow CLI brings that contract to AI-powered education workflows. It gives agents and developers a practical way to create and evaluate learning content without treating the terminal as an afterthought.
For teams building learning products, internal academies, or AI-assisted training operations, TutorFlow can become part of the workflow instead of another tool someone has to manually operate.
Start with the CLI Overview
If your team is exploring how AI agents can create and evaluate learning content, start with the CLI command surface before building a custom integration.
The TutorFlow CLI overview walks through installation, authentication, JSON output, and the main commands for agent-ready education workflows.
Use it to verify the workflow, test real commands, and decide where TutorFlow CLI fits in your content generation, assessment, or training automation stack.
FAQ
What is TutorFlow CLI?
TutorFlow CLI is the command-line interface for the TutorFlow Content API and the Agent Platform account routes. It lets developers, AI agents, and scripts create learning content and run AI evaluations from the terminal.
What can TutorFlow CLI generate?
It can create multi-lesson courses, slide decks, interactive modules, tests, narrated videos, games, and simulations, and it can grade learner answers with AI evaluation.
Can TutorFlow CLI run in CI or agent environments?
Yes. It supports environment-variable authentication, JSON output, non-interactive inputs, structured errors, reliable exit codes, and --wait for jobs that run in the background.
Which API key does TutorFlow CLI use?
A Content API key that starts with tf_content_. tutorflow agent register creates an account and saves one; a test key (tf_content_test_) works in a sandbox classroom and spends nothing. Older tf_platform_ keys still work for account and billing commands, but not for content commands.
