openapi-cli-gen 0.0.17

Generate typed Python CLIs from OpenAPI specs with Pydantic model flattening into CLI flags

openapi-cli-gen

Generate a full CLI from any OpenAPI spec in seconds. Nested request bodies become flat --flags automatically.

# Instead of this:
curl -X POST /api/users -d '{"name": "John", "address": {"city": "NYC", "state": "NY"}}'

# You get this:
mycli users create --name John --address.city NYC --address.state NY

Install

# Recommended: pipx (installs in isolated environment)
pipx install openapi-cli-gen

# Or with uv
uv tool install openapi-cli-gen

# Or in a virtual environment
pip install openapi-cli-gen

Try It Now

Point at any public API — no setup, no files needed:

# Get a random cat fact
openapi-cli-gen run --spec https://catfact.ninja/docs --base-url https://catfact.ninja Facts get-random

# Browse cat breeds as a table
openapi-cli-gen run --spec https://catfact.ninja/docs --base-url https://catfact.ninja Breeds get --limit 5 --output-format table

┏━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ breed          ┃ country       ┃ origin         ┃ coat       ┃ pattern       ┃
┡━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ Abyssinian     │ Ethiopia      │ Natural/Stand… │ Short      │ Ticked        │
│ Aegean         │ Greece        │ Natural/Stand… │ Semi-long  │ Bi- or tri-…  │
│ American Curl  │ United States │ Mutation       │ Short/Long │ All           │
└────────────────┴───────────────┴────────────────┴────────────┴───────────────┘

# Inspect any spec to see what commands you'd get
openapi-cli-gen inspect --spec https://petstore3.swagger.io/api/v3/openapi.json

Generate Your Own CLI

openapi-cli-gen generate --spec https://api.example.com/openapi.json --name mycli
cd mycli && pip install -e .
mycli users list
mycli users create --name John --email [e-mail] --address.city NYC

Ship it to your users: pip install mycli.

Nested Model Flattening

The core feature. Works at any depth:

--address.city NYC                                  # depth 1
--ceo.name Bob --ceo.email [e-mail]             # depth 2
--retry.backoff.strategy exponential                 # depth 3
--tags admin --tags reviewer                         # arrays
--environment JAVA_HOME=/usr/lib/jvm                 # dicts
--role admin                                         # enums (validated)
--address '{"street": "123 Main", "city": "NYC"}'   # JSON fallback

As a Library

from openapi_cli_gen import build_cli

app = build_cli(spec="openapi.yaml", name="mycli")
app()

Or plug API commands into your existing CLI:

from openapi_cli_gen import build_command_group

registry = build_command_group(spec="openapi.yaml", name="mycli")

Auth

Auto-configures from your spec's securitySchemes:

export MYCLI_TOKEN=sk-xxx    # env var
mycli users list --token sk-xxx  # or flag (overrides env)

Pre-Built CLIs

We publish ready-to-use CLI wrappers for popular APIs, generated with this tool. Install one and start using it instantly:

# Full-coverage CLI for the OpenAI REST API
pipx install openai-rest-cli
export OPENAI_REST_CLI_TOKEN=sk-...
openai-rest-cli Chat create-completion --model gpt-4o-mini --messages '[{"role":"user","content":"Hi"}]'

openai-rest-cli — every OpenAI endpoint (chat, embeddings, images, moderations, files, vector stores, batch) exposed as a typed command. PyPI

# Full-coverage CLI for Meilisearch
pipx install meilisearch-rest-cli
meilisearch-rest-cli Health get

meilisearch-rest-cli — every Meilisearch REST endpoint, generated from their official OpenAPI spec. PyPI

More CLIs coming: Qdrant, Typesense, Airflow.

Tested Against Real APIs

36/36 regression tests passing across 6 live APIs. Full CRUD validated — not just reads.

API	Type	Tests	Notes
OpenAI	AI/LLM	8/8	Models, Chat Completions, Embeddings, Images (DALL-E), Moderations, Files, Vector Stores
Qdrant	Vector DB	14/14	Collections + Points CRUD, semantic search with real similarity scores
Meilisearch	Search	7/7	Health, version, indexes, documents, tasks, stats
Typesense	Search	1/1	Manually verified, health works live
GitHub	Public	6/6	Meta, licenses, users, rate limit, zen, octocat
Apache Airflow 3.2.0	Workflow	26/27	Full CRUD: create/patch/delete connections, trigger DAG runs

Real commands that work today:

# OpenAI Chat — one command, real GPT-4o-mini response
openapi-cli-gen run --spec <openai-spec> Chat create-completion \
  --model gpt-4o-mini \
  --messages '[{"role":"user","content":"Hello"}]'

# Qdrant vector search — create collection, insert vectors, semantic search
openapi-cli-gen run --spec <qdrant-spec> --base-url http://localhost:6333 \
  Collections create --collection-name pets --vectors '{"size": 4, "distance": "Cosine"}'

openapi-cli-gen run --spec <qdrant-spec> --base-url http://localhost:6333 \
  Search query-points --collection-name pets --query '[0.1, 0.2, 0.3, 0.4]' --limit 5

# Airflow — trigger a DAG with datetime params
openapi-cli-gen run --spec <airflow-spec> --base-url http://localhost:28080 \
  DagRun trigger-dag-run --dag-id my_dag --logical-date 2026-04-09T12:00:00+00:00

# GitHub — public API, no auth needed
openapi-cli-gen run --spec <github-spec> --base-url https://api.github.com \
  users users/get-by-username --username torvalds

See CHANGELOG.md for the full list of bug fixes and improvements.

Compared to Alternatives

Feature	openapi-cli-gen	specli	restish	Stainless
Nested model flattening	All depths	Scalars only	No	2 levels
Generates distributable code	Yes	No	No	Yes
Runtime mode (no codegen)	Yes	Yes	Yes	No
Pluggable into existing CLI	Yes	No	No	No
Open source	Yes	Yes	Yes	No

Status

Early release. Core features work. Issues and feedback welcome.

License

MIT