# Context in. Plans out.

Connectors do two jobs: bring your context in, and hand the finished plan off to the issue tracker where your dev team already works.

A connector is a link between Plansmith and one other tool. Code connectors let the planning agent read a repository so Specs fit the product as it works today, read access only. Issue-tracker connectors push a finished Plan out as issues. Push is one way.

## Context in: what the agent reads

### Code: the product as it works today

- [GitHub](/connectors/github): Repositories, files, and code search. Read access only.
- [GitLab](/connectors/gitlab): GitLab.com or self-managed. Read access only.
- [Bitbucket](/connectors/bitbucket): Bitbucket Cloud and Data Center. Read access only.

Read access only. The agent reads your product's code so the plan fits reality; it never changes a line.

### Wiki: what your team already wrote down

- [Confluence](/connectors/confluence): Your team's wiki, in the planning conversation. Both directions.

The agent searches your spaces, reads a page and its comments, and writes back when you approve. Every write asks first.

### MCP: everything else your company knows

- [Any MCP source](/connectors/mcp): MCP is the open standard that links AI tools to other software. Connect a source that speaks it and the planning agent reads what your company already knows: Notion, Google Drive, your internal wiki. A connector is the bridge to one tool. MCP is the bridge to the rest.

## Handed off to: where your dev team works

### Project management: where plans land

- [Linear](/connectors/linear): Nested issues, estimates, and build order in one push.
- [Jira](/connectors/jira): Issues, sub-tasks, and story points, already filled in. Both directions.
- [Asana](/connectors/asana): Tasks and subtasks with dependencies already set.
- [Radial](/connectors/radial): The issue tracker where dev teams build with agents.

One click pushes the whole plan to your issue tracker. Handed off, not handed back.

## No tracker? Still works.

Export any Plan as a document (Markdown) or a spreadsheet (CSV) and hand it off however your team works. Every Spec keeps its acceptance criteria, its effort size, and its place in the build order. Connect a tracker whenever you are ready.

## Questions, answered

### What can Plansmith connect to?

Three kinds of thing. For context: GitHub, GitLab, and Bitbucket repositories, plus documents, web pages, and any source that speaks MCP. For your wiki: Confluence, which the agent reads and, with your approval, writes back to. For handoff: Linear, Jira, Asana, and Radial. Context comes in, a finished Plan goes out.

### Do I need a connector to get value?

No. Plan, prototype, and export from day one: any Plan becomes a document (Markdown) or a spreadsheet (CSV). Connect a tracker when you want the one-click push, and a repository when you want the Specs grounded in your code.

### Does Plansmith write to my code?

No. Code connectors are read only. The agent lists files, reads them, and searches the default branch, and that is the entire surface. It never writes, commits, branches, or opens a pull request.

### What happens if I push the same plan twice?

Push is one way, and Plansmith remembers which issue belongs to which Spec. A re-push updates the issues for Specs that changed and creates issues for new ones. Issues for unchanged Specs are left alone, and issues your team created themselves are never touched.