inference.sh

Run 150+ AI apps in the cloud with a simple CLI. No GPU required.

inference.sh

Install CLI

curl -fsSL https://cli.inference.sh | sh
infsh login

What does the installer do? The install script detects your OS and architecture, downloads the correct binary from dist.inference.sh, verifies its SHA-256 checksum, and places it in your PATH. That's it — no elevated permissions, no background processes, no telemetry. If you have cosign installed, the installer also verifies the Sigstore signature automatically.

Manual install (if you prefer not to pipe to sh):

# Download the binary and checksums
curl -LO https://dist.inference.sh/cli/checksums.txt
curl -LO $(curl -fsSL https://dist.inference.sh/cli/manifest.json | grep -o '"url":"[^"]*"' | grep $(uname -s | tr A-Z a-z)-$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/') | head -1 | cut -d'"' -f4)
# Verify checksum
sha256sum -c checksums.txt --ignore-missing
# Extract and install
tar -xzf inferencesh-cli-*.tar.gz
mv inferencesh-cli-* ~/.local/bin/inferencesh

Quick Examples

# Generate an image
infsh app run falai/flux-dev-lora --input '{"prompt": "a cat astronaut"}'

# Generate a video
infsh app run google/veo-3-1-fast --input '{"prompt": "drone over mountains"}'

# Call Claude
infsh app run openrouter/claude-sonnet-45 --input '{"prompt": "Explain quantum computing"}'

# Web search
infsh app run tavily/search-assistant --input '{"query": "latest AI news"}'

# Post to Twitter
infsh app run x/post-tweet --input '{"text": "Hello from AI!"}'

# Generate 3D model
infsh app run infsh/rodin-3d-generator --input '{"prompt": "a wooden chair"}'

Local File Uploads

The CLI automatically uploads local files when you provide a path instead of a URL:

# Upscale a local image
infsh app run falai/topaz-image-upscaler --input '{"image": "/path/to/photo.jpg", "upscale_factor": 2}'

# Image-to-video from local file
infsh app run falai/wan-2-5-i2v --input '{"image": "./my-image.png", "prompt": "make it move"}'

# Avatar with local audio and image
infsh app run bytedance/omnihuman-1-5 --input '{"audio": "/path/to/speech.mp3", "image": "/path/to/face.jpg"}'

# Post tweet with local media
infsh app run x/post-create --input '{"text": "Check this out!", "media": "./screenshot.png"}'

Commands

Task	Command
List all apps	infsh app list
Search apps	infsh app list --search "flux"
Filter by category	infsh app list --category image
Get app details	infsh app get google/veo-3-1-fast
Generate sample input	infsh app sample google/veo-3-1-fast --save input.json
Run app	infsh app run google/veo-3-1-fast --input input.json
Run without waiting	infsh app run <app> --input input.json --no-wait
Check task status	infsh task get <task-id>

What's Available

Category	Examples
Image	FLUX, Gemini 3 Pro, Grok Imagine, Seedream 4.5, Reve, Topaz Upscaler
Video	Veo 3.1, Seedance 1.5, Wan 2.5, OmniHuman, Fabric, HunyuanVideo Foley
LLMs	Claude Opus/Sonnet/Haiku, Gemini 3 Pro, Kimi K2, GLM-4, any OpenRouter model
Search	Tavily Search, Tavily Extract, Exa Search, Exa Answer, Exa Extract
3D	Rodin 3D Generator
Twitter/X	post-tweet, post-create, dm-send, user-follow, post-like, post-retweet
Utilities	Media merger, caption videos, image stitching, audio extraction

Related Skills

# Image generation (FLUX, Gemini, Grok, Seedream)
npx skills add inference-sh/skills@ai-image-generation

# Video generation (Veo, Seedance, Wan, OmniHuman)
npx skills add inference-sh/skills@ai-video-generation

# LLMs (Claude, Gemini, Kimi, GLM via OpenRouter)
npx skills add inference-sh/skills@llm-models

# Web search (Tavily, Exa)
npx skills add inference-sh/skills@web-search

# AI avatars & lipsync (OmniHuman, Fabric, PixVerse)
npx skills add inference-sh/skills@ai-avatar-video

# Twitter/X automation
npx skills add inference-sh/skills@twitter-automation

# Model-specific
npx skills add inference-sh/skills@flux-image
npx skills add inference-sh/skills@google-veo

# Utilities
npx skills add inference-sh/skills@image-upscaling
npx skills add inference-sh/skills@background-removal

Reference Files

• Authentication & Setup
• Discovering Apps
• Running Apps
• CLI Reference

Documentation

• Agent Skills Overview - The open standard for AI capabilities
• Getting Started - Introduction to inference.sh
• What is inference.sh? - Platform overview
• Apps Overview - Understanding the app ecosystem
• CLI Setup - Installing the CLI
• Workflows vs Agents - When to use each
• Why Agent Runtimes Matter - Runtime benefits

Workflow

1. Discover — infsh app list --search "<keyword>" or infsh app list --category <image|video|llm|search|3d|x|utility> to find the right app
2. Inspect — infsh app get <app-id> for description, pricing, input schema, and rate limits
3. Generate sample input — infsh app sample <app-id> --save input.json to get a valid input template pre-filled with defaults
4. Customize — edit the JSON input file: replace prompts, file paths, or configuration values
5. Run — infsh app run <app-id> --input input.json for synchronous execution, or --no-wait for long-running tasks
6. Check status — infsh task get <task-id> to poll progress of async tasks; infsh task logs <task-id> for streaming output
7. Retrieve output — outputs are URLs (images, videos) or text responses; download with infsh task download <task-id>

For multi-step pipelines, chain apps: run the first app, extract its output URL, feed it as input to the next app (e.g., generate image → upscale, or generate video → add audio).

Error Handling

Error	Cause	Fix
infsh: command not found	CLI not installed or not in PATH	Run `curl -fsSL https://cli.inference.sh
401 Unauthorized	Not logged in or token expired	Run infsh login and re-authenticate
402 Payment Required	Account has no credits or payment method	Check infsh billing; add credits via inference.sh dashboard
429 Too Many Requests	Rate limit exceeded for the app or account tier	Wait for the Retry-After header duration; for paid tiers, request a limit increase
400 Bad Request on file upload	File path does not exist, is a directory, or exceeds size limits	Verify the file path with ls -la; check the app's max input size in infsh app get <app-id>
Task timed out (HTTP 504 or task status failed)	App execution exceeded max duration	Check infsh task get <task-id> for error details; retry with smaller input or check app status page
502 Bad Gateway or connection errors	inference.sh backend or model provider is down	Check https://status.inference.sh; retry with exponential backoff (2s, 4s, 8s)
Output URL returns 404	Generated asset expired (files are ephemeral)	Re-run the job; for permanent storage, download immediately with infsh task download <task-id>
JSON parse error on --input	Input file is not valid JSON or schema mismatch	Use infsh app sample <app-id> --save input.json to regenerate a valid template
Image/video generation produces blank or corrupted output	Model inference failed silently	Re-run the job; check app-specific known issues in infsh app get <app-id>; try a different model for the same task

Checklist

• App is authenticated before running (infsh login check)
• Model supports the requested input format (image/video/text sizes)
• Task ID saved for async operations to retrieve results later
• Credit cost checked before bulk operations
• Output format confirmed (URL, file, or stdout) before passing to next step

Sources

• inference.sh CLI documentation — installation, commands, authentication
• inference.sh Apps Catalog — full app library with schemas and examples
• OpenRouter API — LLM model routing and pricing
• FLUX documentation (fal.ai, replicate.com) — image generation parameters and best practices
• Veo API (Google DeepMind) — video generation capabilities and limitations
• Tavily Search API (tavily.com) — web search and extraction endpoints
• Exa API (exa.ai) — semantic search and content extraction
• x.com Developer API — Twitter/X posting, DMs, and engagement endpoints

Anti-Patterns

Anti-pattern	Why it fails	Fix
Not checking infsh app get before running	Input schema varies between apps; guessing fields causes 400 errors	Always run infsh app sample first to see the exact schema
Hardcoding file paths without verifying they exist	The CLI auto-uploads relative paths; missing files produce confusing errors	Use absolute paths or verify relative paths with Test-Path before running
Not using --no-wait for long-running tasks	Synchronous runs block the terminal for minutes on video/3D generation	Use --no-wait and poll with infsh task get <id> for tasks expected to take > 30s
Ignoring rate limits and credit costs	Each app has different pricing; some cost $0.50+ per run	Check infsh app get <app-id> for cost estimate before bulk/generating runs
Using the cheapest model for everything	OpenRouter free models may have lower quality or stricter rate limits	Match model quality to task importance; paid tiers are more reliable
Not saving task IDs	Async tasks complete after the terminal session; without the ID you can't retrieve results	Log all task IDs: infsh task get <id> > task-<id>-result.json
Running image generation without checking supported resolutions/aspect ratios	Models silently crop or stretch to supported dimensions	Check infsh app get for supported_sizes or aspect_ratios field