Belt
Install belt so your coding agent can run 250+ hosted AI apps, persist knowledge, search skills, and wire MCP from one small CLI.
Overview
belt is an agent skill for the Build phase (also Ship and Operate) that exposes the belt CLI to run 250+ cloud AI apps, manage typed agent knowledge, search skills, and connect MCP servers.
Install
npx skills add https://github.com/belt-sh/cli --skill beltWhat is this skill?
- Single ~4MB belt binary with curl install and belt login for cloud auth
- belt app search/get/sample/run for 250+ apps (Flux, Veo, Tavily, ElevenLabs, upscalers)
- belt know semantic search, typed memory (skill, concept, observation, reference, preference), create from files or stdin
- belt skill search and featured store for reusable agent workflows
- Bash(belt *) scoped tool access for safe agent invocation
- 250+ AI apps
- ~4MB single binary
- 5 belt know types (skill, concept, observation, reference, preference)
Adoption & trust: 1.5k installs on skills.sh; 3 GitHub stars; 1/3 security scanners passed (skills.sh audits).
What problem does it solve?
You want your agent to call many AI services and remember context without maintaining separate SDKs, auth flows, and ad-hoc scripts for each provider.
Who is it for?
Indie builders shipping agent-heavy products who want one CLI for models, memory, and skill discovery instead of bespoke integrations.
Skip if: Teams that forbid shell network access, need fully offline inference only, or already standardized on a different agent platform CLI without belt.
When should I use this skill?
Use the belt CLI — run 250+ AI apps, manage knowledge, search skills, connect MCP servers
What do I get? / Deliverables
Your agent can discover, sample, and run hosted apps, read and write belt know entries, and browse the skill registry from predictable Bash commands.
- Executed app run outputs (files or JSON)
- belt know entries created or updated
- Skill registry search results for follow-on workflows
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
belt is cataloged under Build because its main job is wiring external models, memory, and registries into the product you are building. Integrations is the canonical shelf for a single-binary CLI that connects agents to cloud apps, MCP, and reusable skill packages.
Where it fits
Run fal/flux-pro with belt app sample and belt app run to generate marketing assets inside the repo workflow.
Smoke-test tavily/search or exa/search outputs before wiring them into production agent tools.
Pipe a post-incident lesson into belt know create as an observation and belt know search it on the next debug session.
How it compares
Use as a unified agent integration CLI rather than pasting one-off curl snippets per vendor into chat.
Common Questions / FAQ
Who is belt for?
belt is for solo and indie builders using Claude Code, Cursor, or similar agents who need run, remember, and discover commands for cloud AI apps and registries.
When should I use belt?
Use belt while building integrations (belt app run), when shipping checks need real model output, and during operations when you belt know create observations or search prior concepts across sessions.
Is belt safe to install?
belt grants Bash access to belt * and cloud login; review the Security Audits panel on this Prism page and your org policy before enabling network-backed CLI tools.
SKILL.md
READMESKILL.md - Belt
## belt cli belt is the cloud platform cli for ai agents. single ~4mb binary, no runtime dependencies. ### install ```bash curl -fsSL cli.inference.sh | sh belt login belt me ``` ### apps — run 250+ ai models ```bash belt app search "flux" # find apps belt app get fal/flux-pro # view schema belt app sample fal/flux-pro --save in.json # generate sample input belt app run fal/flux-pro --input in.json # run it belt app run fal/flux-pro --input '{"prompt": "..."}' --save output.png ``` common apps: - image: `fal/flux-pro`, `fal/real-esrgan` (upscale) - video: `google/veo-2`, `seedance/seedance-2-i2v` - search: `tavily/search`, `exa/search` - audio: `elevenlabs/tts` ### knowledge — persistent agent memory ```bash belt know search "query" # semantic search belt know list --type observation # list by type belt know get namespace/name # get details belt know create ./file.md --type concept # create from file echo "learned X" | belt know create - --name x --type observation # from stdin belt know delete <id> ``` types: `skill`, `concept`, `observation`, `reference`, `preference` ### skills — reusable workflows ```bash belt skill search "deployment" # search registry belt skill store --featured # browse featured # use on-demand (no install, stdout) belt skill use namespace/skill-name # from store belt skill use github.com/user/repo # from github belt skill use user/repo --skill name # pick from multi-skill repo # install persistently belt skill add namespace/skill-name # auto-detects agents belt skill add ns/name --agent claude-code belt skill list # list installed belt skill upload ./my-skill # publish ``` ### connectors — mcp servers ```bash belt mcp list # available connectors belt mcp search "slack" # search belt mcp connect slack # connect belt mcp tools slack # list tools belt mcp run slack send_message --input '{"channel": "#general", "text": "hello"}' ``` ### suggest — unified search ```bash belt suggest "how to generate images" # searches apps + skills + knowledge ``` ### tips - use `--json` for structured output when piping - use `--save filename` to save media outputs directly - `belt app sample` generates valid input — start there for unfamiliar apps - `belt update` to get latest version