Questions we actually get asked

Associates AI Teammates is a platform for deploying AI agent coworkers ("Teammates") into your business. Each Teammate runs on a persistent agent server with a real filesystem, real Docker, and real access to the tools your work lives in. You pick the models you want, mix providers, and your configuration, memory, and track record stay with you. Small teams share a single agent server across roles. Engineering teams give each role its own server so Product, Engineering, and QA Teammates can collaborate privately on code.

Sign up and enter your card — you'll have full platform access and one agent server immediately, with ~600 models to pick from. Configure your first Teammate, point it at the channels you want (Slack, Telegram, email, etc.), and start talking to it. Most first Teammates are up and running the same day.

No. Out-of-the-box Teammates (SEO, Sales, BOS, Marketing) are configured through the dashboard and talk to you through Slack, Telegram, email, or the web. The Engineering Team Teammates exist for teams that want AI to write code — if that's not you, ignore them.

Every flagship model on the same day it's released — Claude, GPT, Gemini, and the rest — plus 600+ other models across providers like Anthropic, OpenAI, MiniMax, OpenRouter, and Venice. Each Teammate picks its own model from the set you've approved, and you can change the allowed list per Teammate. A one-click "Recommended Models" preset covers the common case, and you tune the allowed list from there. (Venice and OpenRouter are providers, not models — they each give access to dozens of model families on their own.)

Yes. Teammates integrate over MCP, webhooks, and direct APIs. Slack, Telegram, Discord, Signal, WhatsApp, and email are built-in channels. Anything else reachable via webhook or MCP (databases, CRMs, search APIs, code forges, monitoring platforms) your Teammates can talk to. Many skills in the skills library ship with common integrations pre-wired.

A Teammate is a configurable AI agent role that lives on one of your agent servers. Each Teammate has a prompt, a model (or set of allowed models), a set of skills it can use, and a personality. You hire Teammates for jobs — "the SEO Teammate," "the Sales Teammate" — and you can reshape or retire them at will.

An agent server is a persistent workspace — a server with a real OS, filesystem, full Docker, and real network access. Teammates live there and do work. Unlike ephemeral containers, agent servers stay on: your repos stay cloned, your databases stay hot, your background processes keep running. When a Teammate picks up work after a pause, the state it left is still there.

Yes. A single server can host multiple Teammates who share a filesystem and can collaborate directly. A shared ops server commonly hosts BOS + SEO + Sales Teammates that the whole human team talks to. A private engineering server commonly hosts Product + Engineering + QA Teammates that collaborate on code with each other but not with anyone else.

If they share a server, they see each other's files and can hand off tasks directly. Across servers, they message via webhooks or the platform API. Deliberately designed: private work stays private, cross-team handoffs are explicit.

Type /model in the chat. That's it. No dashboard required, no reconfiguration — the next message uses the new model. The dashboard is there too if you prefer to set per-Teammate defaults, but day-to-day switching is a slash command.

No — and this surprises a lot of people. The platform gives you access to every model through our own provider accounts. One subscription, one invoice, every model. You never touch Anthropic's billing, OpenAI's rate limits, or Google's quota forms. Model usage is billed at the provider's list rates and shows up on your monthly statement.

Yes. Starting from scratch or from an existing Teammate template, you configure the prompt, model, skills, and channels. You can also write your own skills — reusable capabilities that any Teammate can pick up. Skills are portable: write one and add it to as many Teammates as you like.

Seat + compute + model usage, billed monthly. Pro Solo is $150/month plus model usage — one operator, one agent server, full platform. Pro Team is $50/seat/month plus agent servers at published platform rates plus model usage at provider list rates. Enterprise is custom. Pricing page has the full breakdown.

Not right now — you start on a paid plan (Pro Solo or Pro Team) and can cancel anytime, month-to-month.

Yes. Month-to-month on every tier. No long-term contracts on Pro Solo or Pro Team. Enterprise terms are negotiated per-deal.

Each Teammate can have multiple inbound webhook endpoints, each with its own auth (bearer, custom header, HMAC), its own event filter, and its own dynamic prompt template that interpolates payload data. Instead of sending a blob of JSON, a webhook message reads like "Review PR #1234: Add rate limiting to auth endpoint" — the Teammate acts on context, not noise.

Model Context Protocol — a standard way for AI agents to talk to tools and services. The platform exposes an MCP server so your Teammates can manage the platform itself: create servers, deploy other Teammates, manage secrets, or talk to other Teammates programmatically. If you want a meta-Teammate that scales your team as demand changes, MCP is how it does that.

We take prompt injection seriously — and we design assuming it will happen. Three things matter in combination: (1) Infrastructure isolation: each team runs on its own dedicated workstation. Nothing a Teammate does can reach another customer's data. Provider API keys live in a managed secrets system, not in plaintext. (2) Least-privilege credentials: API keys you give your Teammates get scoped to exactly the permissions you want — a Teammate with only-read access to your CRM can't write to it even if an injected message begs it to. Reduce scope per skill. (3) Read-only lifecycle for sensitive files: skill definitions, personality files, config files, and secret stores are kept read-only at the filesystem layer through the Teammate's lifecycle. Even a fully-compromised agent conversation can't mutate them. The config-change-with-human-approval flow (Teammate sends a pending update, you deploy it) is how safe changes happen. Webhooks support HMAC for origin verification. Audit logs track every config deploy. Enterprise adds SSO, custom compute isolation, and audit-log exports.