SoloHQ turns each business unit's repos, docs, and tools into a harness agents run in, and keeps it current on its own. Skills, context, tasks, and the swarm, so you can run as many agents as the work needs.
A skill is a plain file: the steps, the context it reads, and a fast check. SoloHQ watches how every run uses it. When runs start missing for the same reason, it rewrites the skill, replays those runs, and ships the new version inside your lanes.
Pages, threads, and code come in from every source and are checked against each other. When two say different things, SoloHQ picks the one the evidence supports and rewrites the file. It asks you only when the evidence is unclear.
One line of intent becomes a plan. Tasks are split where steps are independent, routed to a model by size, and verified on the way out. The ones that touch money, auth, or customer data wait for a person. Nobody drags cards.
Run as many agents as the work needs. SoloHQ sets the caps, picks models by task size, stops retries at the source, and reports cost per verified task, not tokens. Bring Claude Code, Codex, Dots, Cursor, or your own; they all read the same harness.
Every task runs in its own sandbox with only the repos, files, and credentials it needs, and the sandbox is destroyed after verification. Secrets never enter the context. Every read, write, and merge is logged against a task, a skill version, and a lane, so an auditor can replay what any agent did and why.
Work-management tools were built for people to plan in an interface, with agents added on top. In SoloHQ the harness sizes and hands out the work. You see a short list of calls that need judgment.
Giving everyone a chat is a good first step. SoloHQ is for the step after it, when agents do real work on your code and knowledge and cost, context, and coordination start to matter.
The labs will keep shipping better agents and runtimes. Good. Each one still needs your skills, current context, sized tasks, and a budget. Build that harness once, and every agent you bring runs in it.
Connect its repos, docs, tools, and agent logs, and let SoloHQ build the harness. Run whichever agents you like in it, and extend it across the enterprise, or the portfolio, when you like what it does.