// developers
DEVELOPERS · MCP + CLI + REST

MCP-first. Zero-token CLI. REST when you need it.

Pipeline ships a hosted MCP server for AI clients, a deterministic CLI for scripts and CI, and a small REST surface for prospect + campaign ingestion. Same API key across all three, 84 tools available to MCP and CLI.

pipeline-gtm·live
$ npx pipeline-gtm scout-run --brief ./tomorrow.md
✓ Sourced 187 prospects from LinkedIn Sales Nav
✓ Deduped against past prospects (162 new)
→ Handed off to researcher
$ npx pipeline-gtm researcher-status
RES · enriched 41 rows · 7 hot signals
→ Handed off to icp gate
$ npx pipeline-gtm icp-score --threshold 70
32 qualified · 9 below threshold → nurture
same key · MCP / CLI / REST84 tools
// mcp · pipeline-mcp

Hosted MCP server. 84 tools.

The recommended way to drive Pipeline from an AI client. Use the hosted endpoint over HTTPS, or run the pipeline-mcp npm package locally as a STDIO proxy.

MCPhosted endpoint
recommended
// claude code · ~/.claude/mcp.json
{
"mcpServers": {
"pipeline": {
"url": "https://www.pipeline.help/
api/mcp-server",
"headers": {
"Authorization": "Bearer <key>"
}
}
}
}
 
// or one-shot
$ claude mcp add pipeline \
https://www.pipeline.help/api/mcp-server
endpoint · pipeline.help/api/mcp-server
MCPlocal stdio proxy
pipeline-mcp
# local STDIO if your client requires it
# (e.g. Claude Desktop without HTTP support)
$ npx pipeline-mcp
 
# claude desktop · claude_desktop_config.json
{
"mcpServers": {
"pipeline": {
"command": "npx",
"args": ["pipeline-mcp"],
"env": {
"PIPELINE_API_KEY": "<key>"
}
}
}
}
npm · pipeline-mcp
tool catalog12 of 84 shown
GA · 84 tools
pipeline:list-agents
pipeline:scout-run
pipeline:researcher-enrich
pipeline:icp-score
pipeline:writer-draft
pipeline:sender-ship
pipeline:closer-reply
pipeline:ghostwriter-draft
pipeline:publisher-schedule
pipeline:watcher-signals
pipeline:engager-comment
pipeline:coach-outcomes
// cli · pipeline-gtm

Zero-token CLI. Mirrors the MCP surface.

Deterministic, cron-safe, pipe-friendly. No model round-trip — same tools, but you call them directly. Use it from CI, shell scripts, or a local cron.

CLIpipeline-gtm
node 20+
# auth once with the same key as MCP
$ export PIPELINE_API_KEY=sk_...
 
# discover commands (mirrors MCP tool names)
$ npx pipeline-gtm --help
 
# run a tool directly
$ npx pipeline-gtm list-agents
$ npx pipeline-gtm scout-run --brief ./brief.md
$ npx pipeline-gtm icp-score --threshold 70
 
# pipe-friendly · JSON out · exit code on failure
$ npx pipeline-gtm export-table tbl_a1 \
--filter "icp>70" --json \
| jq -r '.rows[].linkedin_url'
 
# cron-safe — exit 0 on success, non-zero otherwise
$ npx pipeline-gtm researcher-enrich \
--campaign cam_8x4f --quiet
npm · pipeline-gtm
// rest · pipeline.help/api

REST endpoints for ingestion.

A small, focused REST surface. Use the MCP server or CLI for the full 84-tool surface; use REST when you want to wire Pipeline into a script that can curl.

RESTbearer auth
v1
# Bearer auth · same key as MCP / CLI
# Generate one from workspace → settings → API keys
 
# POST /api/v1/prospects — bulk-add prospects to a campaign
$ curl https://www.pipeline.help/api/v1/prospects \
-H "Authorization: Bearer $PIPELINE_KEY" \
-H "Content-Type: application/json" \
-d '{
"campaign_id": "cam_8x4f",
"prospects": [{
"linkedin_url": "linkedin.com/in/jane",
"first_name": "Jane",
"company_name": "Helix"
}]
}'
 
# GET /api/campaigns — list + create campaigns
# GET /api/auth/validate — verify the key is live
base · pipeline.help/api

// note: The REST surface is intentionally narrow. For full coverage (research, scoring, drafting, sending), use the MCP server or pipeline-gtm CLI — they expose the same 84 tools.

// use cases

What teams build.

▸ USE CASE
Drive Pipeline from Claude Code

Connect the hosted MCP server in your AI coding tool. Run campaigns, inspect rows, and iterate on prompts from your agent loop.

claude mcp add pipeline https://www.pipeline.help/api/mcp-server
▸ USE CASE
Trigger workflows from CI

Wrap pipeline-gtm in a GitHub Action. Kick off a research run, push enriched rows into a campaign, fail the build on a contract violation. Deterministic exit codes, no model in the loop.

$ npx pipeline-gtm scout-run --brief ./brief.md
▸ USE CASE
Bulk-ingest prospects from your stack

POST a payload of LinkedIn URLs (or full prospect rows) into an existing campaign. Same Bearer token, rate-limited per workspace.

POST /api/v1/prospects · Authorization: Bearer …
BUILD ON PIPELINE

One key. Three surfaces.

$99 / seat · BYOK · cancel anytime