LOGIKASign in
On this page

Teams and environments

One person with one app needs none of this: a project, one environment, and their own agent. As soon as there's a staging environment, a colleague, or a second repository, a few things need saying: whose data is whose, who gets emailed, and whose agent starts when something breaks. The rule throughout is that each person acts as themselves, and nobody's agent starts on someone else's behalf.

Projects

A project is what you pay for and what holds the data: its ledgers, its environments, its members and its settings. Each paid project is priced on its own, with its own subscription, payment method and billing portal, which any of its owners can open. Several repositories can belong to one project, each with its own logika.json and its own ledger, and a rule in one can expect an event of another. Every new account also gets one free demo project, for one person.

Environments

An environment exists because an SDK key named it. Creating a key for a name the project doesn't have yet creates that environment:

logika api keys create '{"ledger": "shop", "env": "prod"}'
logika api keys create '{"ledger": "shop", "env": "staging"}'

The key decides the environment, never an attribute your app sets, so a staging key can't write production data. logika init creates none, and none is special: there's no production marker, and prod is just what you called it. A key made without a name is for default, and an owner can also add one by name in the console's Settings, set up as another (an editor, with logika api env add).

Each environment is apart from the others:

  • its own issues: a failure in staging is staging's, and never mixes with production's;
  • its own data, stored apart and deleted on its own;
  • its own days of history: the tier's, or fewer, so staging can keep a week while production keeps a month;
  • its own wake setting and its own subscribers, below.

Renaming one keeps its history, issues and subscriptions. Deleting one is refused while a live key still writes to it.

Roles and seats

Roles are per project.

RoleCan
Ownereverything an editor can, plus the plan and billing, inviting and removing people and changing roles, the project's name, each environment's name, days and agent settings, deleting the project
Editorsign in the CLI and their own agent (a token), wake that agent, create app keys and add environments, choose their own emails and spending limit
Viewersee everything, and choose which emails they get; no token, no agent
TierSeats
Free demo, Solo1
Teamup to 5
Businessunlimited

Inviting starts on Team. An invite works once, for 7 days, and only for someone signed in with the invited address verified. Members and pending invites both take seats, so moving a project down a tier waits until both fit.

Who gets which email

Email is per member and per environment: each member is subscribed to the environments they want to hear about. When an owner invites someone they choose which (every environment, unless they say otherwise); afterwards each member changes their own, and nobody else can. Whoever creates the project is subscribed to every environment as it's created.

  • New issues in an environment go to its subscribers, once.
  • The agent's report goes to the subscribers of the environments its issues came from, in the same thread.
  • A project's notice (a missed payment, a cancellation) goes to its owners.

Only members are ever emailed. Unsubscribing stops all of Logika's email to that address; to get fewer, follow fewer environments. Emails shows what each says.

Whose agent wakes

Every editor or owner who runs logika login in the repository has their own Logika Monitor, signed in as them, which starts their own coding agent on their own subscription, with their own permissions. So:

  • Wake my agent, in the email, the console or the monitor's menu, starts your agent, never a colleague's. With monitors on several machines, the next of yours to connect takes it.
  • In an environment set to auto, the first member's monitor to claim the new issues takes them, and the email tells that member their agent is on it.
  • What an agent does is emailed to its own member alone: you never get an email about a teammate's agent.

Each environment's wake is set by the project's owners in the console's Settings, under Environments, with the rest of what the agent does there: which agent runs (Claude Code, Codex, or a command of your own), how it delivers a fix (a pull request, a deploy, or instructions of your own) and how long a run goes. A run works one environment, with its settings. on_request is the default for every environment, so nothing works on its own until someone sets it; a new environment can copy another's settings.

One agent on the issues at a time

Exactly one agent works the issues at a time, whoever's it is: it claims the open ones, takes in whatever is raised while it works, and closes them all with one report, so two agents can never race to fix the same thing. Wake my agent and Fix it still work while it runs: what they ask for is taken in by the running agent, or by the next one once it finishes.

A Why? someone presses on an insight in Analytics is never refused: each starts a session of its own beside that run, which answers that one question and changes no code, however many are asked at once.

Spending limits

Every run costs whatever your agent's subscription charges for it, so each member sets a soft monthly limit for their own runs in the project, across all its environments, in the console's settings, where everyone's limit shows. Nobody else can change yours; your agent changes it only when you ask.

Past the limit, your monitor starts nothing more in that project until the month turns or you raise it; a run already going finishes as usual. You're told once that month, and your emails say so instead of offering Wake my agent. Until you set one there is no limit at all, and there is never a hard one. A colleague's limit is theirs: reaching yours never stops their agent.