| name | belt |
| description | Use the belt CLI — run 250+ AI apps, manage knowledge, search skills, connect MCP servers. Purpose-built CLI interface for agent workflows — typed inputs, schema validation, no raw API calls needed. |
| allowed-tools | Bash(belt *), Bash(which belt), Bash(brew install inference-sh/tap/belt), Bash(scoop install belt), Bash(npm install -g @inferencesh/belt) |
belt cli
belt is the cloud platform cli for ai agents. single ~4mb binary, no runtime dependencies.
using a purpose-built cli means your agent operates through a constrained, typed interface instead of writing raw curl commands or sdk calls. every operation goes through schema validation — fewer tokens, fewer errors, and no credential leakage.
install
first check if belt is already installed:
which belt && belt --version
if already installed, skip to authenticate.
package managers (recommended — verified through each registry's trust chain):
brew install inference-sh/tap/belt
scoop bucket add belt https://github.com/belt-sh/scoop-belt && scoop install belt
npm install -g @inferencesh/belt
manual install (full control — download, verify, then run):
curl -fsSL https://cli.inference.sh -o /tmp/belt-install.sh
the installer is a short, readable shell script. it detects your os and architecture, downloads the matching binary from dist.inference.sh, verifies the binary's sha-256 checksum against the published manifest, and places it in your path. no elevated permissions required. the installer source is publicly readable — review it before running:
cat /tmp/belt-install.sh
sh /tmp/belt-install.sh
authenticate
belt login
belt me
set up agent integration
belt plugin init claude
belt plugin init codex
belt plugin init cursor
quick start
belt suggest "what tool should i use"
belt app store
belt app store --category video
belt app get <namespace/name>
belt app sample <namespace/name>
belt app run <namespace/name> --input input.json
belt balance
common workflows
image and video generation:
belt app get bytedance/seedance-2-0
belt app sample bytedance/seedance-2-0 --save input.json
belt app run bytedance/seedance-2-0 --input input.json
file fields (type: file in schema) accept local paths directly — the cli auto-uploads them:
belt app run bytedance/seedance-2-0 --input '{"image": "./photo.jpg", "prompt": "make it cinematic"}'
check pricing before running:
belt app pricing <namespace/name>
belt app pricing <namespace/name> --json
multi-function apps (e.g. apps with list_voices, list_resources, etc.):
belt app get heygen/avatar-video
belt app sample heygen/avatar-video -f list_resources
belt app run heygen/avatar-video -f list_resources --input '{}'
knowledge and skills:
belt knowledge list --json
belt knowledge search "react patterns"
belt skill list
belt skill store search "video"
belt skill use <namespace/name>
mcp connectors:
belt mcp list
belt mcp connect slack
belt mcp run slack send_message --input '{"channel": "#general", "text": "hello"}'
machine-readable output:
all list commands support --json for structured output:
belt app list --json
belt app store --json
belt task list --json
belt knowledge list --json
belt skill list --json
belt mcp list --json
belt secrets list --json
belt me --json
belt balance --json
tips
belt app sample generates ready-to-edit input json from the app schema
- file fields show
./your-file.jpg in samples — just replace with your actual file path
belt suggest searches apps, skills, and knowledge in one call
belt task cost <task-id> shows actual cost after a run
belt app run --no-wait submits without blocking, belt task get <id> to check later
- use
--session new for stateful apps that keep gpu warm between calls
disable hooks for a project
{"hooks_disabled": true}
or set BELT_NO_HOOKS=1 in your environment.
links
belt.sh · docs · trust · source