LaunchBuddy for your AI agent

What is blocking this launch, what should happen next, and can LaunchBuddy make it?

Install one router for Claude, Codex, and Cursor. Assess any public launch for free. Connect only when you want saved projects, a ranked growth plan, or approved launch work.

Install LaunchBuddy →Try the web assessment

ClaudeCodexCursorMCP

Install once

Assess before login. Connect when the work should persist.

The installer adds the LaunchBuddy skill and MCP bridge to all three supported clients. Anonymous assessment works immediately; connection is optional until you need a private project.

Terminal

# 1. Install the router skill and MCP bridge
npx getlaunchbuddy install --all

# 2. Start with a free public assessment in your agent
/launchbuddy assess https://yourproduct.com

# 3. Optional: connect for saved projects, plans, and asset work
npx getlaunchbuddy connect

Connection opens LaunchBuddy in your browser for approval. Your agent never sees your LaunchBuddy password.

The agent loop

Diagnosis first. Plan second. Creation only after approval.

The router follows one vertical path so a useful assessment can become durable work without turning login into a paywall or letting an agent spend money on its own.

Assess the public launch

Paste a public URL. LaunchBuddy scores the launch, names the primary gap, shows the strongest evidence, and returns a shareable report plus a short-lived claim token. No account is required.

Free · anonymous

Connect and claim the original scan

A browser-approved device connection attaches the agent to your LaunchBuddy account. The claim token saves that exact assessment as a project without scanning the site or paying for the audit twice.

Free · no card

Build the growth plan

LaunchBuddy converts the assessment into a ranked plan of launch work, separating assets it can prepare from advisory channel actions.

Included when connected

Preview the highest-leverage asset

The agent shows the exact title, brief, evidence, and expected impact before asking you to approve creation.

Preview before approval

Create through explicit review

An eligible connected account gets one no-card trial asset, bound to the top recommendation and a single workspace card. The trial starts here—not at login—and rendering starts only after web approval. Paid projects use the same review path.

One trial asset · human approval

Reassess and prove the lift

After approved work ships, reassess the same project to compare the score, close the original gap, and choose the next best action.

Before / after loop

Identity is not billing

The handoff between identifying a user and paying is explicit.

Installation, identity, product entitlement, and Stripe payment are four separate states. Existing paid users keep their access after connecting; free users keep their reports and plans without subscribing.

Identify

Connect for continuity, not conversion.

Browser approval links the device to an account, then a short-lived claim token attaches the original anonymous scan. The agent lists only owned projects and resolves the exact project before discussing entitlement.

Save and retrieve assessments
Claim the original scan without rerunning it
Create the assessment-derived growth plan
Preview the recommended asset
Reassess after changes ship

Pay only for execution

Show the work and the offer before checkout.

After you approve an exact asset preview, LaunchBuddy creates a non-charging intent. An eligible free account receives one scoped trial asset. Once it is consumed, the agent shows the returned offer and asks separately before opening checkout.

Existing paid project: skip checkout
Eligible connected project: one no-card trial asset
Free project: show the exact $49/month offer
Stripe webhook activates the same resumable intent

State	What the agent can do	Payment
Anonymous	Assess a public URL, share the report, and receive a claim token	No
Connected free	Claim/save projects, create the plan, preview assets, and reassess	No
Single-asset trial	Create and render one assessment-recommended asset through workspace review	No card; one per account or domain
Paid active	Creation, ongoing growth, and workspace publishing eligibility	Stripe active
Paid lapsed	Read prior work and reassess; creation returns a restart handoff	Restart required

Safe by default

The agent can prepare the decision. You control the external action.

No surprise checkout

Connecting never creates a checkout, trial, subscription, or charge.

Bounded free usage

Scans are rate-limited, cached, and protected by a durable daily budget. The trial is one card and one render.

No silent publication

The MCP creation call does not render or publish. Workspace review remains required.

No duplicate purchase

Access and prior trial/paid history are checked against the account, domain, and exact owned project before checkout is presented.

Test the public flow

You can verify the distribution surface in five minutes.

Start anonymous, connect, then confirm the agent shows the plan and preview before it reaches any commercial handoff.

Install: run npx getlaunchbuddy install --all.
Assess: ask /launchbuddy assess https://launchbuddy.com and confirm you receive a score, primary gap, evidence, and report URL.
Connect and claim: run npx getlaunchbuddy connect, approve the code in your browser, then ask the agent to continue. Confirm it claims the prior scan instead of running a second assessment.
Plan and preview: ask for the growth plan, then ask to preview the top recommended asset. Confirm the agent shows the exact brief before asking to create it.
Check the boundary: the first eligible creation should return one reviewable trial proposal; rendering begins only after workspace approval. A later creation should return the exact offer. Neither path may charge or publish automatically.

Developer check: inspect the public MCP tool contract

curl -s https://launchbuddy.com/api/mcp \
  -H 'content-type: application/json' \
  --data '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Reference

One small CLI. Ten focused MCP tools.

The CLI handles installation, browser-approved identity, and the local bridge. The MCP surface keeps the agent on the assessment-to-work path instead of exposing the entire LaunchBuddy application.

CLI commands

install --all: Install the router skill and MCP bridge.
connect: Approve a free device connection in the browser.
login: Alias for connect.
status: Show the local connection state.
logout: Remove local LaunchBuddy credentials.
mcp: Run the local stdio-to-hosted MCP bridge.

MCP tools

assess_launch: Score a public launch and return the durable report.
get_assessment: Read a saved assessment by scan ID.
claim_assessment: Save the original anonymous assessment after connection without rerunning it.
list_projects: List projects owned by the connected account.
create_growth_plan: Create or read the cached assessment-derived plan.
preview_recommended_asset: Show the exact highest-priority asset brief without spend.
create_recommended_asset: Create or resume the non-charging review intent.
get_action_intent: Read entitlement, approval, render, and verification state.
get_access: Resolve trial and paid capabilities for an exact project.
reassess_launch: Refresh the same project after changes and report the score delta.

Quick troubleshooting

Command not visible: restart the agent client after installation so it reloads skills and MCP configuration.
Private tool says authentication required: run npx getlaunchbuddy connect.
Your paid access is missing: confirm the browser connection uses the same account, then list projects and check the exact project ID.
The device code expired: rerun connect to generate a fresh ten-minute code.

Questions

Straight answers about access and control.

Does connecting start checkout or a subscription?

No. Connecting establishes identity and project continuity only. It never creates a checkout, product trial, subscription, provider job, or publication.

When does LaunchBuddy ask for payment?

Not at install, connection, claim, planning, or preview. An eligible connected account can make one recommended trial asset. After that is used—or when the account or domain already used its trial—the non-charging intent returns the exact offer and the agent asks separately before opening checkout.

What if I already pay for LaunchBuddy?

The agent lists your owned projects and checks access against the exact project. A paid-active project skips checkout and continues to the reviewable workspace proposal.

Can the agent publish something without me?

No. This version exposes no MCP publish tool. Creation becomes a proposed workspace card, and rendering or publishing stays inside the explicit LaunchBuddy review flow.

Where are my agent credentials stored?

The CLI stores short-lived access and rotating refresh credentials locally in your LaunchBuddy configuration directory with user-only file permissions. Server records contain hashes, not raw tokens.

Start with the diagnosis

Find the launch gap before you make another asset.

Run the public assessment now, or install LaunchBuddy where you already work.

Run a free launch scan →View install commands