Pipeline

A local-first, AI-native operating system that lets one person sell, deliver, and run a business from the terminal.

Quick Start

git clone https://github.com/your-org/pipeline-cli.git
cd pipeline-cli
npm install
npm run build
pipeline init

Usage

# Add a contact and a deal
pipeline contact:add "Jane Smith" --email jane@acme.com --org "Acme Corp"
pipeline deal:add "Acme Consulting" --contact jane --value 15000 --stage proposal

# Ask the agent who needs a follow-up
pipeline agent "who should I follow up with this week?"

# Review and approve the agent's proposed actions
pipeline approve

# Explore the relationship graph
pipeline related jane

# Check your pipeline at a glance
pipeline status
pipeline timeline --days 30

# Close a deal and start delivery
pipeline deal:close acme --won --create-project
pipeline milestone:add "Deliver v1" --project acme --due 2026-04-01
pipeline project:show acme

Commands

Topic	Commands	Description
contact	add list show edit rm tag note	Manage contacts (people + org + CRM metadata)
deal	add list move close note pipeline	Track deals through stages (close --create-project)
org	list show edit contacts	Organizations and their members
task	add list done	To-dos attached to contacts or deals
log	call email meeting list	Record interactions
email	send inbox thread sync	Send and receive email (via Composio Gmail)
agent	(free-form prompt)	Chat with the AI assistant
agent	enrich follow-up digest qualify	Built-in agent workflows
schedule	add list remove run install uninstall	Automated agent runs via cron
project	add list show move close	Delivery projects linked to won deals
milestone	add list done	Milestones within projects
field	get set rm	Custom fields on any entity
config	get set	Read/write configuration
top-level	init status search timeline related link unlink import export approve dashboard	Setup, search, graph, data exchange, approvals

Every command supports --json for scripted output and --help for details.

AI Agents

Pipeline ships with nine built-in agents:

Agent	What it does
agent:enrich	Research a contact and update their records
agent:follow-up	Find stale contacts and propose follow-up emails
agent:digest	Morning pipeline briefing and daily summary
agent:qualify	Assess deal health and suggest next actions
agent:inbox	Review inbox and propose replies to incoming emails
agent:deal-manager	Review all active deals for staleness and propose actions
agent:meeting-prep	Prepare a briefing for an upcoming meeting with a contact
agent:task-automator	Ensure every active deal has appropriate tasks
agent:delivery-check	Review active projects for delivery health and propose actions

Run any agent with a free-form prompt:

pipeline agent "draft an intro email to Sarah about our consulting services"

Agents never act directly. Every mutation (send email, update stage, create task, log note, create relationship edge) is inserted into a pending_actions queue. You review and approve them:

pipeline approve          # interactive review
pipeline approve --all    # approve everything

Set ANTHROPIC_API_KEY in your environment to use agents.

Scheduled Runs

Automate agents on a recurring schedule:

pipeline schedule:add follow-up --cron "0 9 * * 1-5"   # weekdays at 9am
pipeline schedule:list
pipeline schedule:remove <id>

# Install/uninstall the cron job that runs due schedules
pipeline schedule:install
pipeline schedule:uninstall

schedule:run executes all due schedules and is called automatically by the installed cron job.

Configuration

Two-level TOML config with deep merge:

File	Scope
~/.pipeline/config.toml	Global (user-wide)
./.pipeline/config.toml	Project-local (overrides global)

Key settings:

[pipeline]
stages = ["lead", "qualified", "proposal", "negotiation", "closed_won", "closed_lost"]
currency = "USD"

[agent]
model = "claude-sonnet-4-6"
auto_approve = false

[email]
provider = "composio"
from = ""

[delivery]
stages = ["kickoff", "in_progress", "review", "complete"]

[integrations]
composio_api_key = "your-composio-api-key"
enabled = true

Read and write config from the CLI:

pipeline config:get agent.model
pipeline config:set email.from "you@yourdomain.com"

MCP Server

Pipeline exposes its service layer as an MCP server over stdio, so any MCP-compatible client (IDEs, other agents) can interact with your Pipeline data.

node bin/mcp.js --db ~/.pipeline/pipeline.db

Development

npm run build        # compile TypeScript (required before running)
npm run dev          # watch mode
npm test             # run all integration tests (vitest)
npm run lint         # check with Biome
npm run format       # auto-fix with Biome

Architecture, conventions, testing strategy, and schema details are documented in CLAUDE.md.

License

MIT