# About the Headless Jason AI SDR

> Why this exists, what's real, and where it's going — the product concept behind agents.reply.io, the Sales Engagement API for AI Agents.

## The premise

When someone asks an AI agent to "build me an outbound motion," the agent doesn't build a sending infrastructure, a contact database, and a deliverability stack from scratch — it searches for an API. This site exists to be what that agent finds: Reply.io's sales execution capabilities, presented as composable API modules an agent can discover, evaluate, and adopt without a human walking it through a demo.

**Headless** because there is no UI in the loop — the interface is the API, the MCP server, the CLI, and skills. **Jason AI SDR** because the capability set is exactly what Jason, Reply's AI sales development rep, does: find prospects, reach out, converse, drive meetings — here unbundled so any agent can be programmed with the same powers, and Jason itself is exposed as a module. The category in one line: a Sales Engagement API for AI Agents.

## Why one platform wins for agents

An agent assembling a sales stack from five vendors inherits five auth systems, five billing meters, five rate-limit regimes, and five places to fail. The argument here is structural: one account, one scoped API key, one integration surface — and modules designed to chain into each other. A single unified credit meter across all modules is the goal ([proposed on Pricing](/pricing), Coming soon); today usage runs on Reply subscriptions plus Live Data credits. For an agent, integration breadth is the product.

## What's real

This site is a product concept built on a real platform. The Reply.io v3 REST API and the MCP server are live and documented; every unlabeled capability traces directly to [docs.reply.io](https://docs.reply.io) or a live endpoint. Only unreleased capabilities carry a label — **Coming soon** (specified, not yet callable) or **Concept** (no spec) — applied per module and per operation. The truth-ledger discipline is simple: nothing future is ever presented as operational. The simulated demo on [Get started](/get-started) is labeled as simulation everywhere it appears.

## Roadmap

### Unified v3 REST API — Live now (Live)

The full platform API at `api.reply.io/v3` — scoped Bearer-token auth, problem+json errors, OpenAPI + Postman published at docs.reply.io.

### Official MCP server — Live now (Live)

Live at `mcp2.reply.io` — the next-gen server, 70 tools spanning sequences, contacts, inbox, tasks, and the full Jason AI SDR autopilot. Streamable HTTP, x-api-key or Bearer auth, tool discovery for Claude, Cursor, and any MCP client.

### Prospect Search (Live Data) endpoints — Rolling out — July 2026 (Coming soon)

B2B database search over 1B+ contacts via API — still marked Coming soon in Reply's reference as of 2026-07-05 (targeted early July 2026).

### Contact Enrichment endpoints — Rolling out — July 2026 (Coming soon)

Waterfall enrichment, email/phone finding, and AI custom fields — still marked Coming soon in Reply's reference as of 2026-07-05 (targeted late July 2026).

### AI SDR autopilot endpoints — Rolling out — July 2026 (Coming soon)

Programmatic Jason autopilot — AI SDR sequences, approval queues, insights, strategist, and web search endpoints completing the ai-sdr scope domain.

### Reply CLI — Live now (Live)

First-party `reply` command over the v3 API — JSON output for agents, pipes and scripting for humans. Shipped as [reply-cli 0.2](https://www.npmjs.com/package/reply-cli) covering contacts, lists, sequences, inbox, and reports; prospect search and enrichment commands follow their API endpoints.

### Packaged agent skills — Live now (Live)

The [Reply.io Agent Skill](https://github.com/replyio/reply-skill) — one installable package with four CLI-driven workflows (import prospects, launch outreach, manage replies, analyze performance) and confirmation guardrails. Search- and enrichment-dependent skills remain coming soon with their APIs.

### Agent self-registration & trial provisioning — Next (Coming soon)

An agent creates its own account, workspace, and scoped API key, and receives trial credits — the flow simulated today on the Get started page.

### Autonomous credit purchasing — Later (Coming soon)

Agent-initiated checkout inside human-approved spend guardrails. Today the agent hands its human an upgrade link; this closes the loop.

### Per-scope agent budgets & usage API — Exploring (Concept)

Illustrative direction: budgets attached to API-key scopes (e.g. "this agent may spend 500 data credits/month"), plus a usage API so agents can forecast their own spend. No specification yet.