Skip to main content
Glama
README.md
# Astra Ops Hub

Nineteen **runnable** business engines that implement the GPT-6 Astra workflow cases from the beamnxw field guide, plus an MCP server, memory evolution log, and approval policy.

This is not a prompt pack. Each case executes against the Northlamp Commerce world fixture, returns a structured artifact, splits **evidence vs assumption**, and refuses unsafe writes until a human approves.

Repo: https://github.com/ST3ELY1212-code/astra-ops-hub

## Install

```bash
git clone https://github.com/ST3ELY1212-code/astra-ops-hub.git
cd astra-ops-hub
python -m pip install -e .
python -m pytest -q
```

## CLI

```bash
python -m astra_ops.cli list
python -m astra_ops.cli run support_resolver
python -m astra_ops.cli run multi_agent --goal "founder week pack"
python -m astra_ops.cli run-all
python -m astra_ops.cli monitor
python -m astra_ops.cli memory
python -m astra_ops.cli focus
```

## MCP

```json
{
  "mcpServers": {
    "astra-ops-hub": {
      "command": "python",
      "args": ["-m", "astra_ops.mcp_server"],
      "cwd": "/absolute/path/to/astra-ops-hub"
    }
  }
}
```

Tools: list_cases, run_case, run_all, monitor, memory_recent, focus_directives.

Open site/index.html for the operator console.

TDQS

C2.7/5.0

Scored across 6 tools

Disambiguation4/5

Each tool has a largely distinct role: listing cases, running one vs. all engines, checking status, reading memory, and reading directives. The only mild overlap is between run_all and run_case, but their descriptions make the scope difference clear.

Naming Consistency3/5

All names use snake_case, but the conventions are mixed: list_cases and run_case are verb_noun, run_all is verb_adverb, monitor is a bare verb, while memory_recent and focus_directives are noun phrases. Readable but not a predictable pattern.

Tool Count5/5

Six tools is well-scoped for an ops hub covering listing, execution, monitoring, and memory/directive reads. Each tool earns its place without bloat or excessive thinness.

Completeness3/5

Core execution and inspection paths exist, but the surface has notable gaps: monitor exposes open approval gates yet there is no approve/reject tool, and there is no way to inspect a single engine's details or status. Memory and directives are read-only, which may be acceptable but limits lifecycle coverage.

Maintenance

ActivitySlowing
ResponsivenessResponsive