@axisagent/cli
v0.1.4
Published
Local Axis runtime, skill registry, policy validator, and CLI.
Readme
Axis Local
Local runtime package for Axis.
It contains the first usable registry, skill manifest generator, policy schema, scanner, and CLI. This is the part that makes the project more than a frontend catalog.
Commands
npx @axisagent/cli validate
npx @axisagent/cli list
npx @axisagent/cli list --chain Base
npm run axis -- inspect bankr-launch-feed
npx @axisagent/cli install bankr-launch-feed --target .axis/skills
npx @axisagent/cli scan .axis/skills/bankr-launch-feed
npx @axisagent/cli connect --pair AXIS-1234 --api https://api.axisagent.xyz --run-approved
npx @axisagent/cli serve --port 8788Axis Relay
The public dashboard should connect through Axis API, not direct browser-to-localhost permission prompts.
Flow:
- Dashboard creates a pairing code through
POST /v1/pairing/create. - User runs
npx @axisagent/cli connect --pair AXIS-1234 --api https://api.axisagent.xyz --run-approved. - The local connector opens an outbound session to Axis API.
- The connector sends heartbeats, detected agents, capabilities and registry stats.
- When the user clicks
Authorize + send, the connector can run approved Axis install+scan tasks locally. - The dashboard polls the task result and opens the Axis approval ticket.
For local API testing:
npx @axisagent/cli connect --pair AXIS-1234 --api http://127.0.0.1:8790 --onceLocal Connector
The localhost connector remains available for development and fallback flows. When the server is running, the dashboard or relay client can call the local connector before any execution:
POST /v1/installinstalls a skill package and scans its policy.POST /v1/scanscans an existing skill package.POST /v1/ticketscreates a simulation ticket with chain, spend cap, permissions, execution plan and policy checks.POST /v1/tickets/:id/decisionrecords user approval or rejection. Blocked tickets cannot be approved.
Ticket controls are grouped by purpose:
core: chain, spend cap, approval, execution mode and session limitspermission: explicit control for every declared permissionpayment: x402 quote, asset and receipt controlsexecution: wallet execution gatesaudit: secrets and local audit requirements
Safety Rules
- Money-moving skills default to
approval-required. - Critical skills must require approval and declare a spend cap or read-only mode.
- Private keys are never stored in generated packages.
- Install output includes
SKILL.md,policy.json, andREADME.md. - Simulation tickets are persisted under
.axis/ticketsfor local auditability.
