Every tool is defined once in @launchengine-dev/contracts and served two ways: as an MCP tool, and over HTTP. The schemas below are read from that registry, so they match what the server accepts.
Create a key on the agents page, then send it as a bearer token:
Authorization: Bearer le_live_...
A key is scoped to one workspace. Requests are rate limited per key (default 60/minute); exceeding it returns a retryable error.
The MCP server is a thin proxy to the same endpoints, published as @launchengine-dev/mcp. It needs no clone and no local API — the default points at production.
{
"mcpServers": {
"launchengine": {
"command": "npx",
"args": ["-y", "@launchengine-dev/mcp"],
"env": { "LAUNCHENGINE_DEV_API_KEY": "le_live_..." }
}
}
}
Cursor (~/.cursor/mcp.json), Cline and Windsurf take the same mcpServers block. VS Code uses servers in .vscode/mcp.json, and can prompt for the key instead of storing it in the file:
{
"inputs": [
{
"type": "promptString",
"id": "launchengine-key",
"description": "LaunchEngine.dev API key (le_live_...)",
"password": true
}
],
"servers": {
"launchengine": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@launchengine-dev/mcp"],
"env": { "LAUNCHENGINE_DEV_API_KEY": "${input:launchengine-key}" }
}
}
}
Claude Code needs one command instead:
claude mcp add launchengine \
--env LAUNCHENGINE_DEV_API_KEY=le_live_... \
-- npx -y @launchengine-dev/mcp
Running npx -y @launchengine-dev/mcp by hand prints one line and then waits — that is a working stdio server, not a hang. It answers --help and --version if you want to check the install.
POST /api/tools/<tool_name>
Content-Type: application/json
{ ...tool input... }
The response is always { ok: true, data } on success or { ok: false, error: { code, message, retryable } } on failure, using the codes in the shared contract.
List the projects in the workspace: name, URL, description, target ICP and competitors. Free. Call this first - every other tool takes a project_id or a signal id that starts here.
No parameters. Send an empty JSON object.
Return the workspace's credit balance and plan. Free; call it before a metered tool (find_intent_signals, draft_reply cost 1 credit each).
No parameters. Send an empty JSON object.
Return scored posts from Reddit, Hacker News and the open web where someone is asking for a product like the project's - highest intent score (1-100) first, with the matched query and the reasons for the score. Reads signals found by earlier scans; pass scan: true to search live sources first. Costs 1 credit.
| Field | Type | Default | ||
|---|---|---|---|---|
| project_id | string | yes | — | Project to search. Get ids from list_projects. |
| Required. Project to search. Get ids from list_projects. | ||||
| min_intent_score | number | no | 60 | Lowest intent score to return, 1-100. 90+ is asking for exactly this now, 75-89 is comparing options, 60-74 is related; below 60 is mostly noise. |
| Lowest intent score to return, 1-100. 90+ is asking for exactly this now, 75-89 is comparing options, 60-74 is related; below 60 is mostly noise. | ||||
| platforms | array | no | ["hackernews","reddit","web"] | Sources to include. "web" is pages found through search indexes. "twitter" is accepted but not scanned yet, so it returns nothing. |
| Sources to include. "web" is pages found through search indexes. "twitter" is accepted but not scanned yet, so it returns nothing. | ||||
| since_days | number | no | 30 | Only posts from the last N days, 1-365. |
| Only posts from the last N days, 1-365. | ||||
| status | enum | no | — | Only signals in this state. Omit for all states. |
| Only signals in this state. Omit for all states. | ||||
| limit | number | no | 20 | Maximum signals to return, 1-100, highest score first. With scan: true it also caps results fetched per search query. |
| Maximum signals to return, 1-100, highest score first. With scan: true it also caps results fetched per search query. | ||||
| scan | boolean | no | false | Refresh from live sources before returning. Slower, and spends one of the workspace's daily scans (free plan: one a day across all projects) on top of the call's credit. Leave false to read signals found by earlier scans. |
| Refresh from live sources before returning. Slower, and spends one of the workspace's daily scans (free plan: one a day across all projects) on top of the call's credit. Leave false to read signals found by earlier scans. | ||||
| web_mode | enum | no | "single" | Only used with scan: true. "single" queries one web-search provider; "parallel" queries every configured provider and merges the results, for wider coverage at more upstream cost. |
| Only used with scan: true. "single" queries one web-search provider; "parallel" queries every configured provider and merges the results, for wider coverage at more upstream cost. | ||||
| include_muted | boolean | no | false | Include posts from authors and subreddits the workspace has muted. Default false, matching the dashboard. |
| Include posts from authors and subreddits the workspace has muted. Default false, matching the dashboard. | ||||
Fetch one signal by id: the post's title, snippet, URL, author, score, reasons, engagement and any saved draft reply. Free.
| Field | Type | Default | ||
|---|---|---|---|---|
| intent_id | string | yes | — | Signal id, from find_intent_signals. |
| Required. Signal id, from find_intent_signals. | ||||
Draft a reply to a signal's post for a person to review and post. Never posts anything. The draft helps with the poster's problem and names no product. Saved on the signal, replacing any earlier draft. Costs 1 credit.
| Field | Type | Default | ||
|---|---|---|---|---|
| intent_id | string | yes | — | Signal to reply to, from find_intent_signals. |
| Required. Signal to reply to, from find_intent_signals. | ||||
| tone | enum | no | "consultative" | Writing style: "consultative" (helpful, conversational), "direct" (gets to the point), or "technical" (practitioner detail). In every tone the draft helps with the problem and names no product, yours included. |
| Writing style: "consultative" (helpful, conversational), "direct" (gets to the point), or "technical" (practitioner detail). In every tone the draft helps with the problem and names no product, yours included. | ||||
Add a keyword for a project's scans to search, or update one already added. A project with at least one keyword is included in the daily scheduled scan (06:00 UTC). Free.
| Field | Type | Default | ||
|---|---|---|---|---|
| project_id | string | yes | — | Project to monitor for. Get ids from list_projects. |
| Required. Project to monitor for. Get ids from list_projects. | ||||
| keyword | string | yes | — | Phrase to search for in every scan of this project, 2-120 characters, e.g. "alternative to Mixpanel". Subscribing the same keyword again updates it. |
| Required. Phrase to search for in every scan of this project, 2-120 characters, e.g. "alternative to Mixpanel". Subscribing the same keyword again updates it. | ||||
| competitors | array | no | [] | Competitor names to store with this keyword, up to 25; replaces the stored list. Scans currently search the project's own competitor list, not this one. |
| Competitor names to store with this keyword, up to 25; replaces the stored list. Scans currently search the project's own competitor list, not this one. | ||||
Metered tools deduct credits before the work runs and refund on failure. Call get_credit_balance before a metered call if you need to check first — it is free.
Sending an Idempotency-Key header makes a metered call safe to retry: the first call is charged and any repeat with the same key is refused with IDEMPOTENT_REPLAY rather than charged again. Send a new key to run the same call a second time. With no header, every call is charged and runs.
draft_reply always returns requires_human_post: true. Nothing is posted on your behalf, by design — Reddit and Hacker News treat automated self-promotion as spam. generated_by reports whether a model or the deterministic fallback produced the text, so a caller can tell a real generation from a generic one.