What you can use today, and where Teamtrail is going next. Every screenshot is the real app, running a real trail.
A trail leads to a destination. Tickets are the decisions and work along the way, with what blocks what.
See what's done, what's being worked, what's blocked and what's waiting on you, with the critical path and progress at a glance.
Switch from the title. Open questions stay tagged with their trail, so the inbox never loses context.
Your agents start from what you already have, and they remember what they learn.
An agent studies what exists and writes it up: the parts involved, the rules your product follows, what could break and what it's assuming. You approve it, or ask for changes.
Rules, habits, pitfalls and dead ends go in your project's field guide. Each agent gets what matters for its task, with the reason why.
Agents ask, you decide, and the decision is written down. You choose which calls are always yours, and only urgent ones interrupt you.
Questions, approvals, stuck work, failing builds and new plans in one list, ranked by what they hold up and how long agents have waited.
Database changes, logins, billing, deleting data, new dependencies: set what always needs you. Agents can't make those calls on their own; they ask.
Every question from every agent, oldest first, with suggested answers and the conversation right below.
Come back after a break and Scout tells you what changed: work done, work now blocked, what agents decided and what it cost.
Every trail writes down what it assumes. When something turns out wrong, an agent proposes a reroute instead of pushing on. Nothing changes until you approve it.
The agent explains what it found, which assumption broke and exactly what it wants to change. Approve it, or send it back with a note.
New work, cancelled work and paused work are drawn right on the map, before anything moves.
Run several agents at once. Each can work in its own copy of your project, and you hear about it before two of them change the same thing. Teamtrail never merges for you.
Agents say which files they expect to change. If two overlap, you see it in your list right away, and you can put one back in line.
Turn it on once, and every task gets a separate copy of your repo on its own branch. Agents never share a checkout, and you review each branch when the work is done.
You agree on what done means before work starts, and nothing counts as done without proof. The checks come from your repo, so no agent can change what judges its own work.
Before building, an agent writes down what has to be true for the work to count, including your product's rules. When it finishes, every point needs proof.
The checks you already use, from your repo. A failure goes back to the agent with what went wrong.
The agent that did the work never reviews it. When you run Claude and Codex, each reviews the other's work first.
For the work you care most about, the ticket closes only after you approve it in the inbox.
What actually changed, the pull request and its build, with a flag when they don't match what the agent said.
Every call across your trails: the question, the answer, the options considered, the reasoning, who decided and the full conversation.
How your agents deliver across every trail, by vendor, agent and kind of work, with plain notes on what stands out. Shown here on a sample project.
How long each trail took, how often you stepped in and what was redone. Trail stats compares it across trails, by vendor, agent and kind of work.
Trails sync both ways. Changes made offline wait and send in order when you're back.
Small screens open on the inbox. Map, ticket, activity and plan are one tap away.
Each of these keeps you in charge: Teamtrail proposes and ranks, and you decide.
An overview card per repo and one queue of what needs you across all of them, answerable from one place.
Hard calls go to stronger models and routine work to cheaper ones, based on how each did on your past trails.
Your agents run in the cloud, and you steer them from the same Lookout, on any screen.
One command in your project folder, or pick Demo in setup to explore a ready-made trail first.