10x CLI
A terminal coding assistant that chains model tiers into multi-step workflows for review, debugging, refactoring, and testing.
- Source repo
- ntegrals/10x
- Stars
- ★ 1.4k
- Last updated
- 10mo ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 39/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changes
- Cost
- Free software; you pay for model usage
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer reviewing a file or directory from the terminal can run
/review <path>for security, performance, and style analysis. - Not a fit if
- Teams unwilling to send code to cloud models like OpenRouter
- Users who want a GUI instead of a terminal command
- Source review
- 39/100 · Major gaps
What does this agent do, and when should you use it?
10x is a coding-assistant CLI installed globally with npm and run from a shell. It offers an interactive session as well as one-shot prompt execution with -x. Its Superpowers break complex work into steps that can use different model tiers. Projects can supply context through a root-level 10X.md, while additional skills and workflows can be defined in .10x/skills/, .10x/superpowers/, or the corresponding user configuration directories. The documented built-in workflows cover code review, pull-request descriptions, refactoring, debugging, code explanation, and test generation.
Running 10x starts an interactive session, while 10x -x "<prompt>" executes a prompt and exits. /review <path> performs security, performance, and style analysis for a path; /pr generates a PR description from staged or committed changes; /refactor <file> provides analysis and implementation for a refactor; /debug <issue> follows reproduce, analyze, and fix steps; /explain <path> explains code architecture; and /test <file> generates a comprehensive test suite. Custom Superpowers are Markdown files with name and trigger metadata; the shown template passes user input through {{input}}, makes an earlier step’s result available as {{previous}}, and assigns tiers such as fast or smart per step.
- A developer reviewing a file or directory from the terminal can run
/review <path>for security, performance, and style analysis. - An engineer preparing a pull request can run
/prafter staging or committing changes to generate its description. - A maintainer who needs to understand and then modify a source file can use
/refactor <file>for guided analysis and implementation. - A developer investigating a reported defect can use
/debug <issue>to work through reproduction, analysis, and a fix. - A project contributor needing tests for a particular file can run
/test <file>to generate a comprehensive suite.
How do you install or deploy this agent?
In a terminal with npm, run:
npm install -g 10x-cliThen start it with:
10xTo use your own OpenRouter API key, run 10x --byok; the documentation does not specify how that key is configured.
How do you use this agent?
Start an interactive session with 10x. For a one-shot request, use 10x -x "Explain src/app.ts". Select a tier with 10x --model <tier>, where the documented tiers are superfast, fast, and smart; restore a session with 10x --resume <name>. Put project context in a root 10X.md, custom skills in .10x/skills/ and invoke them as /<skill-name>, and custom workflows in .10x/superpowers/.
What are this agent's strengths and limitations?
- It includes concrete commands for review, PR-description generation, refactoring, debugging, explanation, and test generation.
- Superpowers can split a task into multiple steps and assign a different model tier to each step.
- Project- and user-level Markdown skill and workflow directories support team-specific extensions.
- A documented
--byokmode uses an OpenRouter API key supplied by the user.
- The documentation does not state where to configure an OpenRouter API key, how authentication works, or how costs are controlled.
- Its model tiers are named around GPT OSS 20B, Kimi K2 1T, and Claude Opus 4; the source does not document how to add models or providers.
- The “up to 20x faster” statement is a project claim without benchmark methodology, workload conditions, or reproducible results.
- Supported operating systems, Node.js versions, offline behavior, data-handling boundaries, and retry behavior are not documented.
How does this agent compare with similar options?
The README’s feature table compares 10x with Claude Code, Cursor, and GitHub Copilot, positioning 10x around multi-step Superpowers, smart model routing, and BYOK; it does not provide independent performance benchmarks or compatibility detail.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| 10x CLI This agent | 39 · Major gaps | CLIFree + model costs | ★ 1.4k | 10mo ago | TypeScript | — |
| Memdoor Coding Workflows | 64 · Some gaps | CLIFree + model costs | ★ 380 | today | Go | OpenAI API · Claude API |
| Friday Agent Workbench | 73 · Some gaps | Desktop appFreemium | ★ 53 | today | TypeScript | Codex · Claude Code |
| Poolside pool | 49 · Major gaps | CLIFree + model costs | ★ 442 | 2mo ago | — | OpenAI API |
How does FollowAgents rate this agent?
Why each dimension lost points
The core entry point exports PermissionManager, permission rules, and settings interfaces, showing that permission controls exist as entry points. The sample does not show defaults, confirmation flows, or tool execution policy, so least privilege and user confirmation receive only thin support. README mentions BYOK and OpenRouter but does not explain key storage, request data scope, or external-effect handling. No rollback mechanism or clear maintainer/source attribution appears in the sample, so these criteria are deducted.
README describes interactive sessions, model routing, and multi-step workflows, and the source exports corresponding router, session, tool, and workflow modules, offering limited corroboration. The 'up to 20x' and routing performance claims are unsupported by the material, and the package manifest does not list runtime dependencies, leaving availability unclear. The sample provides no failure handling or user-facing failure guidance, so failure_messages scores 0.
README targets coding tasks and lists review, debugging, refactoring, explanation, and test-generation scenarios, with examples for custom workflows, skills, and project context. Core exports show extensible tools, skills, and workflow modules, but platform support, environment constraints, and capability boundaries are sparsely documented. Workflow triggers and CLI commands are concrete, supporting a stronger trigger_precision score.
README is organized into quick start, features, configuration, and CLI sections, with an npm install command and examples, but omits prerequisites, troubleshooting, and known limitations. The MIT LICENSE text is complete and earns full marks. The npm package name, repository badges, and private root package/version metadata do not fully clarify release-version relationships; there is no changelog or clear maintenance responsibility, so those criteria receive partial scores.
Command and workflow examples help users understand intended use, while multi-model steps, skills, and project context suggest potential added value. The sample offers no inspectable real outputs or evidence for cost, speed, or quality trade-offs. The README's 20x speed and model-speed labels are unsupported promotional claims, so effectiveness is scored conservatively.
The README's install command, command list, and configuration examples are directly traceable; the core entry point exports many corresponding modules, providing limited cross-checking. Performance, routing, and comparison claims have no source or methodology in the sample, which also does not distinguish verified facts from promotion or inference. All three criteria therefore receive partial scores.
- The supplied material gives no measurement method or support for the 'up to 20x' speed claim or model speed labels; do not rely on them as a performance expectation.
- The sample does not explain API key storage, what data is sent to model providers, or confirmation and recovery behavior for write and shell tools; review the full implementation and configuration guidance before use.
Maintain this project? Add the FARS badge to your README
[](https://followagents.com/en/agents/10x)<a href="https://followagents.com/en/agents/10x"><img src="https://followagents.com/badge/10x.svg" alt="FARS review"></a>Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
- Pre-alpha version of HyperGen - PyPi package: https://pypi.org/project/hypergen/View on GitHub ↗
FAQ
Do I need to provide an API key?
10x --byok for an OpenRouter API key you provide.Can a team create its own workflows?
.10x/superpowers/ can define a trigger command and a multi-step flow using YAML front matter.What does it produce?
Can I continue a previous session?
10x --resume <name> for resuming a session, but the storage location and retention policy are not documented.Thanks — we use this to improve the page.