Skip to main content
The Reminix CLI isn’t just for deploying — you can invoke agents, call tools, and stream responses directly from the terminal. This makes it a powerful integration path for automation, scripting, and AI agent tool-use.

Invoke agents

Call any deployed agent from the command line.
# One-shot invocation
reminix agent invoke support-bot --prompt "How do I reset my password?"

# Stream the response token by token
reminix agent invoke support-bot --prompt "Summarize this report" --stream

# Multi-turn chat
reminix agent chat support-bot --message "Hello"

# With conversation history
reminix agent chat support-bot \
  --message "What about pricing?" \
  --history '[{"role": "user", "content": "Hello"}, {"role": "assistant", "content": "Hi! How can I help?"}]'

# Structured input
reminix agent invoke analyzer --input '{"data": [1, 2, 3]}' --stream
Use --verbose to see execution ID, status, duration, and a link to the logs dashboard:
reminix agent invoke support-bot --prompt "Hello" -v

Call tools

Call any deployed tool with JSON input.
# Call a tool
reminix tool call lookup_customer -i '{"email": "user@example.com"}'

# Verbose mode shows execution details
reminix tool call create_ticket -i '{"subject": "Bug report", "body": "Login fails"}' -v

# List available tools
reminix tool list

JSON output for scripting

Every command supports -o json for machine-readable output. This makes it easy to pipe results into other commands.
# Get agent response as JSON
reminix agent invoke support-bot --prompt "Hello" -o json | jq '.output'

# List agents as JSON
reminix agent list -o json | jq '.[].name'

# Call a tool and extract a field
reminix tool call lookup_customer -i '{"email": "user@example.com"}' -o json | jq '.id'

Use in Claude Code skills

The CLI works as a tool inside any AI agent that can execute shell commands — including Claude Code. You can create a Claude Code skill that invokes your Reminix agents:
# A Claude Code skill can call your agents
reminix agent invoke code-reviewer \
  --prompt "Review this pull request for security issues" \
  -o json

# Or call tools to look up context
reminix tool call get_customer_context \
  -i '{"customer_id": "cust_123"}' \
  -o json
Any AI coding assistant that supports shell execution (Claude Code, Cursor, Windsurf, GitHub Copilot) can call your deployed agents and tools through the CLI.

Use in CI/CD pipelines

Call agents as steps in your build pipeline. Here’s an example with GitHub Actions:
# .github/workflows/review.yml
name: AI Code Review
on: [pull_request]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Install Reminix CLI
        run: npm install -g @reminix/cli

      - name: Run AI review
        env:
          REMINIX_PROJECT: ${{ secrets.REMINIX_PROJECT }}
        run: |
          DIFF=$(git diff origin/main...HEAD)
          reminix agent invoke code-reviewer \
            --prompt "Review this diff: $DIFF" \
            -o json > review.json

      - name: Post review comment
        run: |
          COMMENT=$(jq -r '.output' review.json)
          gh pr comment ${{ github.event.pull_request.number }} --body "$COMMENT"

Shell scripting

Use the CLI in bash scripts for automation workflows.
#!/bin/bash
# Process a batch of support tickets

for ticket_id in $(cat ticket_ids.txt); do
  echo "Processing ticket $ticket_id..."

  # Look up ticket details
  TICKET=$(reminix tool call get_ticket \
    -i "{\"id\": \"$ticket_id\"}" -o json)

  SUBJECT=$(echo "$TICKET" | jq -r '.subject')

  # Generate a response
  RESPONSE=$(reminix agent invoke support-bot \
    --prompt "Draft a response for: $SUBJECT" -o json)

  echo "$RESPONSE" | jq -r '.output' >> responses.txt
done

Environment variables

Configure the CLI for non-interactive environments:
VariablePurpose
REMINIX_PROJECTDefault project (overrides config file)
REMINIX_API_URLAPI URL (for self-hosted instances)
# Set project for the session
export REMINIX_PROJECT=acme/my-app

# Now all commands use this project
reminix agent list
reminix tool list