Deployment architecture

DevWorkWire is a local Python package exposing the dwire console script. There is no deployed service, listener, or MCP server. An AI agent invokes the same installed command from a local shell. See ADR-011.

Distribution model

The application repository declares a Python 3.12+ package and a dwire console entry point. Installation and release channel details must be checked against the current packaging requirement and proposed ADR-007; this page does not claim PyPI or Homebrew publication has occurred.

The portable skill ships as skills/devworkwire/SKILL.md in the application repository. Users place it in their agent's supported skills directory and ensure the agent's shell can find the installed dwire command. The skill has no separate Jira credentials or runtime.

Runtime environment

Each invocation runs on the user's machine with access to the configured Jira project and, for folder imports, the Markdown and local resume file. Direct commands may return one JSON result on stdout with --format json; logs go to stderr. A CLI upgrade changes the behavior seen by both people and agents.

Data durability and recovery

.devworkwire-import.json beside the source Markdown stores the Jira connection identity, created issue keys, and source hashes. It must be kept with the folder for safe re-runs. If Jira's response to a write is uncertain, inspect Jira before resolving the pending local attempt. There is no SQLite preview store in the current CLI.

Future deployment questions

A future MCP server would need its own client requirement, security review, and distribution design. It is not part of the current package interface.