Agents
An agent is a specialized AI worker. Rather than one general assistant, Erdo gives you a team: one analyzes data and answers questions, one writes and runs code, one builds dashboards, and so on. You brief an agent in plain language and it does the work end to end — connecting to your data, reasoning, and producing a result — instead of just describing how. Agents can use tools (query a dataset, run code, search the web, send an email, place a call) and hand off to one another when a task spans specialities. You don’t pick tools manually; the agent decides what it needs for the job. See Agents.Threads
A thread is a conversation with an agent — your unit of work in Erdo. You ask for something, the agent works, and results (tables, charts, pages, files) appear inline in the thread. Threads keep their full history, so you can come back, pick up where you left off, and see exactly how a result was produced.Datasets and Connectors
Agents work on your data, and there are two ways to give it to them:- Datasets — data that lives in Erdo: files you upload (CSV, Excel) or tables synced from a connected source. Agents query datasets directly to answer questions and build on them.
- Connectors — authorized links to external tools and sources: your databases, Google Workspace, Slack, and many more. Connecting a source lets agents read from it (and, with your approval, act in it). You manage these under Data → Connectors; each connection is per user and encrypted.
Some capabilities depend on a specific connector. For example, an agent can only
book a meeting on a call once you’ve connected Google Calendar — see
Voice.
Knowledge
Knowledge (the Brain in the app) is what your agents learn and remember about your business — the metrics that matter and how they’re defined, the shape of your data, reusable skills, and decisions you’ve made. Agents read from it to stay consistent and write to it as they learn, so the workforce gets smarter over time instead of starting cold every thread. See Knowledge.Review and approvals
Erdo keeps a human in the loop. Agents propose rather than silently change: new knowledge an agent wants to remember shows up in Review for you to accept or reject, and consequential actions (sending an email, writing to a connected system, placing a call) ask for your approval before they run. You stay in control of anything that leaves Erdo or changes your data.Pages
A page is a small app an agent builds for you — a dashboard, report, or tool — wired directly to your data with live values and realtime state. Pages run in a managed sandbox under your existing permissions, and you can share them with a private editor link or a public link. See Pages — and Build Apps for the developer view.Activity and automations
Agents don’t only work while you watch. Under Activity you can set up automations — an agent that runs on a schedule or reacts to an event, so a report, analysis, or alert happens on its own. Longer-running work shows up here too: workstreams track multi-step jobs (a campaign, a lead engine) as agents move them through phases, and experiments run structured tests and record the decision. See Automations.Voice
Erdo agents can place outbound phone calls — qualify a lead, confirm a detail, or book a meeting live on your calendar. See the Voice guide.Put it together
The quickstart connects a data source and walks an agent through a real task.