# Solo dev using Claude Code daily: built a "team workspace" for me and my agents. Looking for honest feedback before going further — retained record

> Retained reddit record from reddit:ClaudeCode.

Canonical: https://highsignal.app/data/records/a1bbe918d3b5c55c

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeCode)  published 2026-10-25 05:12 ingested 2026-10-26 02:39

[original source ↗](https://www.reddit.com/r/ClaudeCode/comments/1xetos3/solo_dev_using_claude_code_daily_built_a_team/)

## retained text



```
***Note:*** *I wrote this and used Claude to help tidy up the wording before posting.*

I've been building most of my software with Claude Code daily for about a year. The coding part got fast. What didn't keep up was knowing where I stood. At any point I couldn't easily answer three questions: what's been done, what's still pending, and what I should do next. Work was spread across chat sessions, branches and notes, and my setup gave me no single view of it.

So I built a "team workspace" for me and my agents: one place where I and the agents see the same picture of the project. It isn't available yet, and I'd like to know whether it's useful to anyone besides me.

**What it does today**

*A board the agents use too*

* Issues, sub-issues, statuses, priorities, projects and cycles: Linear-style.
* Agents connect over MCP (or REST) with their own key, so their changes show up as theirs.
* Each agent gets a "what changed since you last looked" feed. When I move a ticket, comment or change a priority, the agent picks it up next session.

*A living map of the codebase*

* A layered explanation of the system: concepts → services and modules → database tables → files. Agents write most of it; humans read most of it.
* When code changes, the parts of the map that describe it are flagged stale until they're re-described.
* Step-by-step flows ("how signup works"), architecture decisions with their reasons, and a glossary.

*Session start without the re-explaining*

* A Claude Code plugin gives the agent a compact briefing when a session starts: what the project is, what's in flight, recent changes, decisions and gotchas.
* At the end of a session it prompts the agent to update the board and the map.

*Run issues on your own machine, with your own subscription*

* A small daemon on your laptop/pc takes an issue, works on it in an isolated git worktree using your own Claude Code login, and pushes a branch for review. (If you happen to have more than one machines you can technically run a lot of agents and the limit is your LLM subscription tier)
* Commits and pull requests are attributed to you, with the agent as co-author.
* You can watch a run's progress from the issue, stop it, or continue it.

*Chat with your codebase from anywhere*

* From the web app, ask questions about your repo. Your own machine answers, read-only.
* It can read other branches and history ("what changed on development since staging?"), and you can switch models mid-conversation.

**What I'd love to hear**

1. Would something like this fit your agentic workflow, or does your current setup (CLAUDE.md, memory files, Linear/Jira plus MCP) already cover it?
2. Which part would you actually use: the board, the map, session briefings, runs on your own machine, or chat from anywhere?
3. Would you trust an agent-written explanation of your system if it showed you which parts are out of date?
4. Running agents on your own machine with your own subscription, or/and in a cloud sandbox: which would you prefer, and why?
5. What's the most annoying part of working with coding agents day to day that nothing solves for you yet?

There's no link. I'm trying to figure out whether this is worth building out and what to focus on. Brutal honesty welcome.
```



document fetched 2026-10-26 00:37 · canonical https://reddit.com/r/ClaudeCode/comments/1xetos3/solo_dev_using_claude_code_daily_built_a_team
