Trov is shared memory for a software team and its coding agents: docs, decisions, the roadmap, a ticket queue and a feed of what shipped, read and written by people and their agents alike. Agents propose changes; a person confirms what becomes official (Concepts).
1. Sign in
Go to trov.dev and choose Get started. Sign in with GitHub or Google: either one creates your account the first time, and no invitation is needed. Google sign-in takes an address Google has verified.
The first time, you pick a handle (prefilled with your GitHub login) and a color. The handle is how work is attributed to you everywhere, including everything your agent writes. Both can be changed later in Settings › Profile, and Settings › Account links the other provider, so either one signs you in.
2. Create or join an organization
Everything in Trov belongs to an organization: its docs, decisions, tickets, roadmap, feed and members. Each organization's data is its own: in Trov, only its members and the agents they connect can read its content. Trov's operators see an organization's name, members and usage, not its content; the Privacy Policy says who else processes it.
- Join one. An admin invites your GitHub login or email address from Org settings › Members. The invitation shows the next time you open Trov.
- Create one. Anyone signed in can create a Free organization (one they own at a time). Free is for up to 3 people; Pro is bought per seat, at $10 per seat / month. Every plan is all of Trov: plans differ only in their limits (pricing).
A new organization opens on its guided setup, one step at a time and every step skippable: an owner or admin connects a repository, connects their coding agent and invites the team; a member connects their coding agent. Help › Guide links back to it.
3. Connect GitHub
In the guided setup, or later in Org settings › Repositories, choose Connect with GitHub. That installs the Trov GitHub App on the GitHub account that owns your repository (or links an installation that is already there), and you pick the repository from the ones you can read. There is no token to paste and no webhook to add. Connecting is approved on GitHub, so an owner or admin who signed up with Google links a GitHub account first.
With a repository connected:
- the repository's GitHub issues show up as tickets, each linked back to its issue; closing or reopening the issue closes or reopens its ticket, and everything else about it is edited in Trov;
- pull requests, checks, deploys and pushes are captured as they happen, for My Work and the Repo dashboard (deploys, CI, drift and usage);
- an agent connection that follows the repository knows which organization to act in from the repository it is working in.
How many repositories an organization can connect is a plan limit (Free: 1, Pro: 5).
4. Connect your coding agent
Your agent reaches Trov over the Model Context Protocol (MCP). You connect it by signing in to Trov in your browser, once; from then on it acts as you: it sees what you see, and what it writes is recorded as yours.
For Claude Code, install the Trov plugin. It wires up the MCP server and adds the skills that run the loop below:
/plugin marketplace add AndresL230/trov
/plugin install trov@trovThen run /mcp, choose trov, then Authenticate. Your browser opens Trov: choose how the connection picks an organization, then click Allow. The connection is listed in Settings › MCP access under Connected apps, where Revoke disconnects it at once.
One connection covers all your organizations. When you allow it you choose how it picks one: follow the repository you are working in, or the organizations you tick, one at a time. You can change that later in Settings.
Any other MCP client (Claude Desktop and claude.ai, Cursor, VS Code, Codex, Windsurf) connects to the same address, https://trov.dev/mcp: setup to copy for each is in Connect an MCP client.
In a chat (claude.ai on the web, Claude Desktop) Trov is a custom connector. There is no plugin and no repository there, so the Allow page suggests manual: you tick the organizations it may use and where it starts. Connect an MCP client › In a chat, with no repository says what a chat can and cannot do.
5. Your first session
- Start as you always do. Before your agent touches an area the team already knows, it reads what Trov has on it (the
load-contextskill does this by itself, and lists any handoffs waiting for you). - Work. Ask “what’s on my plate?” for your tickets. Your agent can update the tickets assigned to you.
- Say “record this session” when you’re done. It stages what shipped: feed entries, doc changes and decisions. They wait in Triage › Review until a person promotes them.
In a chat there are no skills, so ask in words: “what’s on my plate in Trov?”, “what does Trov say about billing?”, “any handoffs waiting for me?”. The agent reads and stages exactly as above; it cannot read git, so to record shipped work give it the pull request or commit.
Not using Claude Code's plugin? Put the snippet from Connect an MCP client › Tell your agent how to use Trov in your project's AGENTS.md or CLAUDE.md, so any agent follows the same loop.