OptionalAutonomous AI AgentsVersion 1.0.1
Using Blackbox AI CLI with Hermes Agent: Multi-Model Coding Tasks
Delegate coding tasks to the Blackbox AI multi-model CLI.
Written by Neura Market from the official Hermes Agent documentation for Blackbox. Commands, paths, and version numbers are reproduced from the source unchanged.
Read the official documentationBlackbox AI CLI Reference for Hermes Terminal
Delegate coding tasks to Blackbox AI via the Hermes terminal using a multi-model coding agent CLI. This reference covers all capabilities, procedures, parameters, constraints, and failure modes for operating the Blackbox CLI.
When to Use
Use Blackbox CLI when you need to:
- Generate, refactor, or review code using multiple LLMs with automatic best-result selection.
- Perform one-shot code generation or quick scratch work in a temporary directory.
- Run long tasks that benefit from background execution and checkpoint/resume.
- Conduct PR reviews without modifying the working tree.
- Execute independent coding tasks in parallel.
Prerequisites
Before delegating tasks, ensure:
- Node.js 20+ is installed.
- Blackbox CLI is installed globally via
npm install -g @blackbox_ai/blackbox-cli(binary:blackbox). - An API key is obtained from app.blackbox.ai/dashboard.
- The API key is configured via
blackbox configure. - Always use
pty=truein all terminal calls.
Parameters and Commands
Core Parameters
--prompt "task"or-p: Non-interactive one-shot execution with the given task description. Required for one-shot tasks.--resume-checkpoint "tag": Resume from a saved checkpoint tag. Optional.--yoloor-y: Auto-approve all actions and model switches. Optional.--vlm-switch-mode: Image-handling mode. Accepts"once"(switch for current query only),"session"(switch for entire session), or"persist"(stay on current model, no switch). Optional.-c, --checkpointing: Enable checkpointing of file edits. Optional.
CLI Management Commands
blackbox configure: Change settings, providers, and models. Required for initial setup and multi-model configuration.blackbox update: Update the CLI to the latest version.blackbox mcp: Manage MCP servers.blackbox extensions: Manage CLI extensions.blackbox voice: Configure voice input.blackbox shortcut: Configure the 'b' shortcut.
Session Commands
During interactive sessions, use:
/compress: Compress the session context./clear: Clear the session./stats: Display session statistics.Ctrl+C: Interrupt the current operation.
Token Limit
The sessionTokenLimit parameter controls the token limit for a session, set in .blackboxcli/settings.json (default 32000). Example configuration:
{
"sessionTokenLimit": 32000
}
Procedures
One-Shot Task Execution
- Ensure prerequisites are met (Node.js 20+, CLI installed, API key configured,
pty=true). - Run the task with a prompt and working directory. For example, to add JWT authentication:
terminal(command="blackbox --prompt 'Add JWT authentication with refresh tokens to the Express API'", workdir="/path/to/project", pty=true)
For scratch work in a temporary directory:
terminal(command="cd $(mktemp -d) && git init && blackbox --prompt 'Build a REST API for todos with SQLite'", pty=true)
Background Mode for Long Tasks
- Start the task in background with PTY:
# Start in background with PTY
terminal(command="blackbox --prompt 'Refactor the auth module to use OAuth 2.0'", workdir="~/project", background=true, pty=true)
# Returns session_id
# Monitor progress
process(action="poll", session_id="<id>")
process(action="log", session_id="<id>")
# Send input if Blackbox asks a question
process(action="submit", session_id="<id>", data="yes")
# Kill if needed
process(action="kill", session_id="<id>")
- Capture the returned
session_id. - Monitor progress using
process(action="poll", session_id="<id>")orprocess(action="log", session_id="<id>"). - If Blackbox asks a question, send input via
process(action="submit", session_id="<id>", data="yes"). - If needed, kill the session with
process(action="kill", session_id="<id>").
Resume from Checkpoint
- After a task completes, note the checkpoint tag displayed by Blackbox (e.g.,
task-abc123-2026-03-06). - Resume with a follow-up task:
# After a task completes, Blackbox shows a checkpoint tag
# Resume with a follow-up task:
terminal(command="blackbox --resume-checkpoint 'task-abc123-2026-03-06' --prompt 'Now add rate limiting to the endpoints'", workdir="~/project", pty=true)
PR Review (Without Modifying Working Tree)
- Clone the repository to a temporary directory, checkout the PR, and run Blackbox:
terminal(command="REVIEW=$(mktemp -d) && git clone https://github.com/user/repo.git $REVIEW && cd $REVIEW && gh pr checkout 42 && blackbox --prompt 'Review this PR against main. Check for bugs, security issues, and code quality.'", pty=true)
Parallel Work
- Spawn multiple Blackbox instances with
background=truefor independent tasks:
terminal(command="blackbox --prompt 'Fix the login bug'", workdir="/tmp/issue-1", background=true, pty=true)
terminal(command="blackbox --prompt 'Add unit tests for auth'", workdir="/tmp/issue-2", background=true, pty=true)
# Monitor all
process(action="list")
- Monitor all sessions using
process(action="list").
Capabilities
- Interactive and non-interactive sessions: Run tasks interactively or one-shot.
- Checkpointing and resuming: Save and resume tasks using checkpoint tags.
- Background mode: Run long tasks in background with session monitoring (poll, log, submit, kill).
- Multi-model execution: Runs the same task through multiple models and uses a judge to select the best output.
- Vision support: Automatic detection of images and model switching for multimodal analysis.
- MCP server management: Via
blackbox mcp. - CLI extensions management: Via
blackbox extensions. - Voice input configuration: Via
blackbox voice. - Shortcut configuration: Via
blackbox shortcut. - Token limit control: Via
.blackboxcli/settings.json. - Update CLI: Via
blackbox update. - Configure providers, models, and settings: Via
blackbox configure.
Constraints and Caveats
- Always use
pty=true— Blackbox CLI is an interactive terminal app and will hang without a PTY. - Use
workdirto keep the agent focused on the correct directory. - Use
background=truefor long tasks and monitor with theprocesstool. - Do not interfere with running sessions; monitor with
poll/log, do not kill because they are slow. - After completion, check what changed and summarize for the user.
- Blackbox uses a credit-based system; multi-model mode consumes credits faster.
- Verify the
blackboxCLI is installed before attempting delegation. - Multi-model mode: configure multiple providers via
blackbox configureto enable Chairman/judge workflow. - Vision support: Blackbox automatically detects images and can switch to multimodal analysis based on VLM mode.
- Checkpoint tag is displayed after a task completes for use with
--resume-checkpoint.
Failure Modes
- Missing PTY (
pty=true) causes the CLI to hang. - Missing or invalid API key prevents authentication.
- Insufficient credits for multi-model or long tasks.
- Node.js version below 20 causes installation or runtime errors.
- CLI not installed:
blackboxcommand not found. - Session killed prematurely due to impatience (do not kill slow sessions).
workdirnot set leads to agent working in wrong directory.- Background session not monitored may leave orphan processes.
Skills the docs pair this with
More Autonomous AI Agents skills
Antigravity CliOperate the Antigravity CLI (agy): plugins, auth, sandbox.Claude CodeDelegate coding to Claude Code CLI (features, PRs).CodexDelegate coding to OpenAI Codex CLI (features, PRs).Computer UseDrive the desktop in the background without stealing focus.GrokDelegate coding to xAI Grok Build CLI (features, PRs).Hermes AgentUse, configure, theme, extend, and orchestrate Hermes Agent.