# Indie Terminal — agent manifest

status: live · index-backed · human-governed
format: agent-readable (humans can skim)
site: https://indieterminal.com

> Research terminal on a 10K+ startup index.
> Visible pipelines. Human approval before anything ships.

## What this is

Indie Terminal is the operating layer for startup intelligence — not a chat wrapper.
Every workflow runs on cited index rows. Pipelines are visible. Output queues for human approval.

## Auth

keys:     /api-keys
header:   Authorization: Bearer <it_live_…>
scopes:   index read · tool runs · agent connections

Generate keys in-app. Same key works for REST and (soon) MCP.

## Surfaces (pick one entry point)

| surface          | path              | use when                          |
|------------------|-------------------|-----------------------------------|
| terminal         | /terminal         | slash commands on live index      |
| command center   | /command          | pipeline runs · action queue      |
| workspace        | /workspace        | notes · canvas · company brain    |
| api keys         | /api-keys         | slack · notion · hubspot · openai |
| docs             | /docs             | full human + API reference        |
| human landing    | /welcome          | marketing view for humans         |

## Terminal commands (index-backed)

indie /radar <category>       — category movers · medians · growth
indie /price <product>        — comp-backed pricing band
indie /value <startup>        — ARR · SDE · blended valuation
indie /research <topic>     — competitor scan · cited comps
indie /launch gtm             — ICP · wedge · channels · sequence
indie /match acquirers        — ranked buyers · outreach draft

All commands cite index rows. No black-box summaries.

## API v1 (REST)

base: https://indieterminal.com/api/v1

GET  /startups/search?q=&mrr_min=&limit=
GET  /startups/:slug
GET  /categories?q=&sort=&limit=
GET  /signals?limit=&since=
POST /compare  { "slugs": ["…", "…"] }

Authorization: Bearer it_live_… on every request.
Responses include cited startup rows where applicable.

## Workflow pipelines

trigger:  natural language → API call → pipeline on index data
states:   queued → running → needs_approval → completed
rule:     nothing ships without human sign-off

Example:
  POST /v1/tools/pricing-audit
  { "category": "dev-tools", "comps": "index", "approval": "required" }

## MCP (coming soon)

Same auth + schema as REST. v1 endpoints exposed as MCP tools — no re-implementation.

## Integrations (agent connections)

slack     — signal alerts · research digests in channels
notion    — export deliverables · research pages
hubspot   — push leads · ICP insights to CRM
zapier    — trigger on research · signal events
airtable  — sync plans · startup data
openai    — bring your own key for research workflows

Connect at /api-keys

## Guarantees

- every output cites index rows
- pipelines show every step — no opaque chat
- action queue: queued → needs_approval → completed
- a human signs off before anything ships

## Quick start for agents

1. Read this file: GET /agent.md
2. Human docs: GET /docs (browser) or scrape for tool schemas
3. Obtain API key: direct user to /api-keys
4. Run index query: GET /api/v1/startups/search?q=dev-tools
5. Queue workflow: command center or tool POST endpoints
6. Poll until needs_approval — surface to human

## Links

- docs:          /docs
- api keys:      /api-keys
- command:       /command
- human view:    /welcome
- pricing:       /pricing