Research Infrastructure for Autonomous Systems

Your agent doesn't understand humans yet.

UserTold.ai is the research layer — a structured pipeline that turns real user behavior into evidence your agent can act on. We run the interviews, extract the signals, and deliver structured JSON your agent reads through MCP or the CLI.

Start Interviewing

$1 / interview prepaidBYOK requiredIssues in your tracker

Pain point evidence card

"I tried this flow three times and still cannot find where to change billing settings."

Signal type: Struggling moment
Confidence: 0.91
Action: Linked to issue #42
Result: Re-tested after release

The loop your agent runs

You design

Define your research protocol—what to learn, how to interview, which modes (talk, speak, observe).

We interview

UserTold.ai runs the interview: captures screen + voice, STS model speaks with real users, records everything.

We analyze

Auto-transcribe, extract pain point evidence cards, provide source materials for verification.

You act

Evidence arrives via MCP tool calls or CLI JSON output — pushed to GitHub Issues or Linear. Your agent reads it and decides what to build.

We measure

Re-interview on the same pain point. Compare signal rates. Proof your fix worked.

$ usertold --help

UserTold.ai CLI — brings human feedback into the agent's decision loop

USAGE
  $ usertold <group> <command> [options]

COMMANDS
  init              Interactive project setup wizard
  project           Manage projects (create, list, status, snippet)
  study             Manage studies (create, list, update, export, import)
  session           Manage sessions (list, get, reprocess, transcript)
  signal            Extract and manage signals from sessions
  task              Create and manage evidence-backed issues
  screener          Manage screeners (create, activate, configure)
  config            Configure per-project settings (BYOK keys)
  setup             Provider setup (GitHub)
  overview          Project dashboard overview

OPTIONS
  --format json     Machine-readable output for agents
  --yes             Non-interactive mode (no prompts)

Who deploys this

Agents for Founders

Run 10 interviews before your next sprint. Know which pain point to fix, backed by signal rates not gut feel.

Autonomous Product Loops

Your agent calls projects.signal_health, runs interviews overnight, creates GitHub issues by morning.

Engineering Agents

Tasks from evidence not debate. When your agent asks "what should I build?", it has an answer backed by real sessions.

Built for agents. Easy to integrate.

MCP Server

Model Context Protocol — the agent-native interface. Your agent calls tools directly: design studies, trigger sessions, read signals, push tasks. No browser required.

# Agent tool calls

projects.signal_health

signals.list

tasks.create_from_signals

tasks.push

CLI (Non-Interactive)

Scriptable setup and orchestration. Full --json and --yes flags for autonomous pipelines.

$ usertold auth whoami --json

$ usertold init \

--org <your_personal_org_handle> \

--name "My Product" \

--format json --yes

REST API

Full API access with JWT auth for embed and programmatic control.

POST /api/projects

GET /api/signals

POST /api/tasks

Simple pricing for autonomous systems

Platform

$1 / interview

Prepaid credit packs starting at $10 (10 credits). Interview orchestration, signal extraction, issue creation, impact measurement, and dashboard. All included.

Inference

BYOK

Bring your own OpenAI key. Your keys, your bill, no markup. Inference costs go directly to your provider account.

Trust and control

You own your data and can delete sessions.
Keys and provider settings are isolated per project.
Security, privacy, and terms pages are always linked and current.

Security Privacy Terms Docs

FAQ

Prepaid credits at $1 per interview ($10 minimum purchase = 10 credits). Inference costs go to your API provider account via BYOK. Your keys, your bill, no markup.

Your agent deserves better evidence.

Set up a project, embed the screener, run interviews, and push pain points to your tracker in under an hour.

Start Interviewing Read Docs