A product-decision context graph your agent can query.
ProductLobster runs senior-PM work over your product and keeps the result as a typed, queryable record of decisions, callable from Claude Code, Codex, Cursor, or any MCP harness. Your agent brings the context. ProductLobster brings the product judgment, and remembers what you decided.
Real PM work, from your harness. Shipped.
All server-side, against your product's Brain. Read by default; each write scope is an explicit opt-in. Every checkpoint decision is captured with the options you weighed, the one you chose, and why — not buried in a chat log you'll never find again.
- Liveplob context-pack
Pull your product's state into your agent with one call — the curated mental models, product summary, locked checkpoint decisions, and an index of recent runs and documents.
- Liveplob audit
Run the 5-stage product audit server-side, against your product's Brain. Returns a run you can poll.
- Liveplob focused-analysis
Run a single focused analysis — competitive landscape, customer research, or strategic frame — grounded in your Brain.
- Liveplob prototype
Generate a clickable prototype from agent-supplied context plus the current Brain state. Retrieve the full screen set with plob prototype-pack.
- Liveplob observe / plob ingest
Write back to the Brain through the content-quarantine gates: a typed observation, or a working-context bundle the server distills into decision-grade signal.
- Liveplob run-replay
Replay a prior run from any stage, re-executed against the context as it was at that point — not today's.
Connect in three commands.
The CLI is on npm. Install it, authorize the connection, and resolve your workspace. Or add ProductLobster as a native MCP server in Claude Code, Claude Desktop, Cursor, Windsurf, or Cline.
$ npm i -g @productlobster/plob
$ plob login # authorize the connection (OAuth)
$ plob workspaces # resolve your workspaceAgents obtain the workspace slug from plob workspaces — never from you. Run plob guide for the full command reference.
A database of the present can't tell you why you got here.
Your PM tool stores the current state of your product. ProductLobster captures the decisionat the moment you make it: the alternatives, the reasoning, the trade-off. Then it lets you replay that thinking later. A database of the present can't tell you why you got here. A record of decisions can.
That's the product-decision context graph: the entities your product already cares about — signals, hypotheses, features, releases — connected by the decisions that moved them and the “why” behind each.
Today this covers decisions made inside ProductLobster, and searchable precedent (“find every call that touched this”) is rolling out. Capturing each decision and its why is real now.
A partner, not autopilot.
ProductLobster doesn't decide for you, and it won't pretend to be right by default. It proposes: here's what we think you decided, here's why, here's the source. You confirm or correct in a click. It surfaces what matters first, ranked by what's at stake, so you're not scrolling through everything it noticed. You stay in control. It just makes sure nothing important gets lost.
The checkpoint flow already works this way. The cross-source review-and-correct queue — surfacing extracted decisions for confirmation, ranked by what's at stake — is coming with the connectors.
From capturing the decision to querying the why.
In active development — framed as roadmap, not present. Join the early-access list and we'll bring you in as it lands.
- Coming soonplob why <decision>
Trace the reasoning, alternatives, and evidence behind any call.
- Coming soonplob precedent
Find every decision that depended on a signal, cited a precedent, or superseded an earlier call.
- Coming soonConnectors
Connect Slack, Linear, and your meeting notes so the graph captures the decisions you make outside ProductLobster too.
Questions builders ask.
- Do I have to move all my work into ProductLobster?
- No. Today you bring your product to ProductLobster. Soon it'll connect to the tools you already use (Linear, your meeting notes, Slack) and build the decision record from there. You keep working where you work.
- Does it make decisions for me?
- No. It captures decisions and surfaces what matters; you make the call and confirm what it captured. A partner, not autopilot.
- How does my agent connect?
- Install the CLI from npm (`npm i -g @productlobster/plob`), run `plob login` to authorize the connection with OAuth, then `plob workspaces` to resolve your workspace. Or add ProductLobster as a native MCP server in Claude Code, Claude Desktop, Cursor, Windsurf, or Cline. Read by default; each write scope is an explicit opt-in.
- Is searchable precedent available today?
- Not yet. Today ProductLobster captures each decision and the why at the checkpoint, and lets you replay a prior run from any stage. Searchable precedent — `plob why`, `plob precedent`, and "what superseded what" — is rolling out next.
Bring the context. We'll bring the judgment.
Start with your productFree during beta. First answers in minutes.