CLI reference

In shortEvery 10xGraph CLI command with its purpose, usage line, options and defaults: init, dev, build, eval, test, skills, audit, config, demo and version.

  • 9 min read
  • 14 sections
  • Updated
  • v1.0.0
  • Markdown

The 10xgraph command ships with the 10xgraph-api package. It has ten subcommands: init, dev, build, eval, test, skills, audit, config, demo and version. Global options go before the subcommand, and -h or --help works everywhere.

Global options

Placed before the subcommand, for example 10xgraph --no-color build.

Option Default Purpose
--format human Output format: human, plain, json or jsonl
--json off Shorthand for --format json
--color auto Color policy: auto, always or never
--no-color off Shorthand for --color never
--progress auto Progress mode: auto, tty, plain, json or quiet
--animation / --no-animation on for an interactive terminal Play the command intro. --no-animation prints a static header instead
--fullscreen / --no-fullscreen off Pin a header and footer on a full-screen surface. TENXGRAPH_FULLSCREEN=1 turns it on
--cwd current directory Run as if started in this directory (must exist)
--verbose, -v 0 Increase diagnostics, repeatable
--quiet, -q off Suppress informational and success output
--debug off Enable debug diagnostics
--yes, -y off Accept recommended defaults
--non-interactive off Never prompt
--version, -V Print the CLI version and exit

--json cannot be combined with a different --format, and --no-color cannot be combined with a different --color.

On an interactive terminal each command opens with a short intro, under a second, that shows the CLI, core and Python versions and then leaves a one-line header in your scrollback. Press any key to skip it, or Ctrl+C to quit at once. Nothing waits for Enter unless you opted into --fullscreen.

Wherever a command takes --config with the default 10xgraph.json, it uses agentflow.json from the same directory when 10xgraph.json is missing. 10xgraph dev flags the old name in its startup output.

init

Scaffold a new agent project. Interactive init asks three questions: the agent name, the setup (Quick or Full) and the AI provider. Everything else comes from the setup.

Terminal
10xgraph init [--path DIR] [--mode quick|full] [options]

Quick writes a minimal config with auth off and no cache_store.

Full writes the recommended setup without asking:

  • "auth": "jwt"
  • "id_generator": "uuid"
  • "cache_store" with "type": "redis" (no url, so Redis comes from REDIS_URL), a rate_limit of 100 requests per 60 seconds by user with fail_open false, and token_revocation on
  • WebSocket and AG-UI off

The summary lists these choices, and the next steps point to 10xgraph config for full control (WebSocket, AG-UI, ID generator, thread naming and every other setting).

The flags are overrides for scripts:

Option Default Purpose
--path, -p . Directory to initialize
--force, -f off Overwrite existing files
--name inferred Agent name, required only when it cannot be inferred
--mode prompt quick or full. --template quick-start|production is an alias
--provider prompt google, openai or anthropic
--auth jwt Full only: none, jwt or custom
--cache-store redis Full only: memory or redis
--rate-limit / --no-rate-limit on Full only: add or leave out cache_store.rate_limit
--token-revocation / --no-token-revocation on with auth Full only: add or leave out cache_store.token_revocation. Ignored with --auth none
--websocket / --no-websocket off Write websocket.enabled
--yes, -y off Accept the defaults without prompting
--non-interactive off Fail instead of prompting
--dry-run off Preview without writing files

Quick ignores --auth, --cache-store, --rate-limit and --token-revocation. Example for a script:

Terminal
10xgraph init --yes --mode full --name shop-support --provider openai --cache-store redis

dev

Start the local development server (Uvicorn) and open the hosted playground when the API is ready. Use --no-open to serve only.

Terminal
10xgraph dev [--config FILE] [--host HOST] [--port PORT] [--open | --no-open]
Option Default Purpose
--config, -c 10xgraph.json Config file
--host, -H 127.0.0.1 Bind address. Use 0.0.0.0 for all interfaces
--port, -p 8000 Port
--reload / --no-reload on Reload on file changes
--open / --no-open on Open the hosted playground when the API is ready
--verbose, -v / --quiet, -q off Logging level

With MODE=production set, dev prints a warning if you also use --reload, or bind to a host other than loopback (127.0.0.1, localhost, ::1): the dev server has no TLS and is not meant for a real deployment.

dev is the only server command. 10xgraph api and 10xgraph play were removed before 1.0. For production, run Gunicorn in the image that 10xgraph build generates.

build

Generate deployment files. See Deploy with Docker and Kubernetes.

Terminal
10xgraph build [--docker-compose] [--k8s] [options]
Option Default Purpose
--output, -o Dockerfile Dockerfile path
--force, -f off Overwrite existing files
--python-version 3.13 Python version of the base image. The minimum is 3.12; an older version is refused
--port, -p 8000 Port to expose
--docker-compose / --no-docker-compose off Also write docker-compose.yml, omit CMD from the Dockerfile
--k8s / --no-k8s off Also write k8s.yaml
--service-name 10xgraph-api Service name in generated files
--worker / --no-worker asked Include a background worker. Needs a non-memory worker key in 10xgraph.json
--scheduler / --no-scheduler off Include the scheduler. Needs the worker
--verbose, -v / --quiet, -q off Logging level

--docker-compose and --k8s cannot be combined: the compose image has no start command, so the pod would crash-loop. The command refuses and exits with an error. Run build once with --k8s, or once with --docker-compose.

At an interactive terminal, build asks where you will deploy, whether to run a worker and the scheduler, and whether to overwrite existing files. See Questions and background jobs.

eval

Run agent evaluations. Discovers *_eval.py and eval_*.py files in the target directory (default evals/) and runs their cases.

Terminal
10xgraph eval [TARGET] [--threshold 0.8] [--parallel]
Option Default Purpose
TARGET evals/ File or directory to evaluate
--output, -o eval_reports Report directory
--no-report off Console summary only
--threshold, -t unset Fail if the overall pass rate is below this value (0.0 to 1.0)
--parallel, -p off Run all cases from all files concurrently
--max-concurrency, -c 4 Concurrent cases when --parallel is set
--open / --no-open open Open the HTML report in your browser when the run ends. Skipped with --no-report, -q, --format json/jsonl, --non-interactive, when CI is set, or when stdout is not a terminal
--verbose, -v / --quiet, -q off Output level

Reports are HTML and JSON unless --no-report is set. When an option is left at its default, the evaluation block of 10xgraph.json can supply directory, output_dir, threshold, parallel and max_concurrency.

test

Run the project tests with pytest.

Terminal
10xgraph test [PATH] [--coverage] [-k EXPR] [-- PYTEST_ARGS]
Option Default Purpose
PATH pytest discovery Test file or directory
--coverage, -C off Run with coverage (--cov=., terminal and HTML reports)
--html off Open the HTML coverage report (needs --coverage)
-k unset Only run tests matching this expression
--verbose, -v / --quiet, -q off Output level

Extra arguments are forwarded to pytest. The test block of 10xgraph.json can set path, coverage and coverage_threshold. An explicit PATH wins over the config.

skills

Install the bundled coding-agent skill, or validate skills.

Terminal
10xgraph skills [--agent codex|claude|github] [--all] [--list]
Option Default Purpose
--agent, -a prompt Target: codex, claude, github, or menu number 1, 2, 3
--all off Install for every supported agent
--list, -l off List supported agents and exit
--path, -p . Project directory to install into
--force, -f off Overwrite an existing installed skill
--validate unset Validate a skill directory or folder of them against the Agent Skills spec. Repeatable
--verbose, -v / --quiet, -q off Logging level

Install locations: .agents/skills/10xgraph plus a 10xGraph section in AGENTS.md (Codex), .claude/skills/10xgraph plus a section in CLAUDE.md (Claude), and .github/skills/10xgraph plus .github/instructions/10xgraph.instructions.md (GitHub).

audit

Read-only environment check: Python version, installed packages, core evaluation API, 10xgraph.json validity, a security check, an audit log check and whether the default port is free. The security check reads only the JSON file and reports production-readiness gaps (auth, rate limiting, token revocation) as warnings; they never fail the command. With MODE=production, the audit log check warns when audit.enabled is not true (see Audit log). Exits 1 if any check fails, so it works as a CI gate.

Terminal
10xgraph audit [--config FILE]
Option Default Purpose
--config, -c 10xgraph.json Config file to validate
--verbose, -v / --quiet, -q off Logging level

revoke

Revoke one token or all of a user’s tokens. It builds the configured cache store the same way the server does (cache_store) and writes to it, so with a shared store the revocation reaches every worker at once and the server needs no restart. Turn the check on with "token_revocation": true inside cache_store in 10xgraph.json. The command refuses with an error when the cache store is InMemoryCacheStore, because a separate process cannot reach another process’s memory.

Terminal
10xgraph revoke (--user USER_ID | --jti JTI) [--before EPOCH] [--expires-at EPOCH] [--redis-url URL] [--config FILE]
Option Default Purpose
--user unset Revoke all tokens of this user_id issued so far (log out everywhere)
--jti unset Revoke the one token with this jti claim
--before unset With --user: revoke tokens issued at or before this epoch time
--expires-at unset With --jti: the token’s exp (epoch seconds), to drop the entry
--redis-url unset Redis URL to write to, in place of the configured cache store
--config, -c 10xgraph.json Config file

Pass exactly one of --user and --jti. With no shared cache store configured (cache_store absent or type memory) and no --redis-url, the command fails.

config

Edit and validate 10xgraph.json in a browser UI. The file is created on first save. If only agentflow.json exists, the editor starts from its contents and saving writes them to 10xgraph.json, leaving agentflow.json untouched.

Terminal
10xgraph config [--config FILE] [--port PORT] [--no-open]
Option Default Purpose
--config, -c 10xgraph.json File to edit
--port, -p 0 (any free port) Editor port
--open / --no-open on Open the default browser

The editor groups every key in a sidebar:

Group Pages
Get running Core
Before production Auth, Cache store (rate limiting, token revocation), Observability
Runs and streaming Runs, WebSocket, AG-UI, Remote tools, Media
Project wiring Checkpointer, ID generation, Thread naming, Memory store, Dependency injection
Developer tools Test, Evaluation

The Cache store page starts with three type cards (Memory, Redis, Custom). Under them, the Rate limiting and Token revocation tabs each have an on/off switch and the fields for that setting. The editor validates with the same parsers as the server and writes a .bak of the previous file on save.

demo

Preview terminal animations without touching the project.

Terminal
10xgraph demo [--style all|typing|network|init|build|eval]
Option Default Purpose
--style all Animation theme

version

Print the CLI and core framework versions.

Terminal
10xgraph version

Accepts --verbose and --quiet.

Exit codes

Commands exit 0 on success and non-zero on failure. Ctrl+C exits 130. 10xgraph test returns the pytest exit code.

Frequently asked questions

What happened to the api and play commands?
Both were removed before 1.0. Use dev, which starts the Uvicorn development server and opens the hosted playground by default; add --no-open to serve only. dev is not meant for production, so use build and Gunicorn there.
Last updated for v1.0.0Edit this page on GitHubReport an issue