Everything Auli does, written down plainly.

01

Create your workspace

Sign up with email, Google, Apple or phone. Every person belongs to at least one workspace, and all data, connections and billing are scoped to it. Invite teammates from Team → Members.

02

Connect a source

Go to Team → Integrations, pick a connector and complete the OAuth handshake. Auli requests read-only scopes and never writes back to your platforms. Tokens are encrypted at rest.

03

Activate the accounts that matter

Agencies and multi-brand teams often see dozens of ad or CRM accounts. Activate only the ones this workspace should read — each activated account counts once against your plan.

04

Ask your first question

Open Auli Chat and ask in plain language: “What changed in pipeline this week, and why?” The analyst plans the query, calls the connectors it needs, and answers with the account and date range attached.

05

Keep what's useful

Pin any chart to a dashboard, export it to PDF or CSV, or share a read-only link. Metric and date-range preferences persist per workspace.

Disconnecting a source revokes the stored token immediately and removes its accounts from the workspace. Historical answers keep their citations but can no longer be refreshed.

Invitations can be sent by email or shared as a link, expire automatically, and are rate-limited. Teams on a verified company domain can allow matching colleagues to join without an invite.

Create a workspace API token in Team → Settings → API access. Tokens are scoped to a single workspace, inherit that workspace’s connected accounts, and are read-only — exactly like the connectors behind them. Every response carries the same citation block the analyst shows in chat.

curl https://api.auli.ai/v1/ask \
  -H "Authorization: Bearer $AULI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "workspace": "acme",
    "question": "What changed in qualified pipeline last week, and why?",
    "date_range": "last_7_days"
  }'

curl "https://api.auli.ai/v1/metrics?metric=conversions&granularity=day&range=last_30_days" \
  -H "Authorization: Bearer $AULI_API_KEY"

{
  "mcpServers": {
    "auli": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.auli.ai/mcp"],
      "env": { "AULI_API_KEY": "sk-auli-..." }
    }
  }
}

MCP tools mirror the analyst’s own toolset: auli.ask, auli.metrics, auli.sources and auli.accounts. Each returns citations alongside the data.

OpenAI — function calling

Register Auli as a tool, then let the model decide when to call it. The tool result already contains the citation chain, so instruct the model to pass citations through verbatim.

from openai import OpenAI
import requests, os

client = OpenAI()

AULI_TOOL = {
    "type": "function",
    "function": {
        "name": "auli_ask",
        "description": "Ask Auli about connected growth data (GA4, Google Ads, "
                       "Meta Ads, LinkedIn, HubSpot, Shopify). Returns a cited answer.",
        "parameters": {
            "type": "object",
            "properties": {
                "question": {"type": "string"},
                "date_range": {"type": "string", "default": "last_7_days"},
            },
            "required": ["question"],
        },
    },
}

def auli_ask(question, date_range="last_7_days"):
    r = requests.post(
        "https://api.auli.ai/v1/ask",
        headers={"Authorization": f"Bearer {os.environ['AULI_API_KEY']}"},
        json={"question": question, "date_range": date_range},
        timeout=60,
    )
    r.raise_for_status()
    return r.json()  # { answer, figures[], citations[] }

resp = client.chat.completions.create(
    model="gpt-4.1",
    tools=[AULI_TOOL],
    messages=[
        {"role": "system", "content": "Use auli_ask for any question about "
                                      "marketing, sales or revenue performance. "
                                      "Always repeat the citations you receive."},
        {"role": "user", "content": "Why did CAC move last week?"},
    ],
)

import anthropic, requests, os

client = anthropic.Anthropic()

tools = [{
    "name": "auli_ask",
    "description": "Ask Auli about connected growth data. Returns a cited answer.",
    "input_schema": {
        "type": "object",
        "properties": {
            "question": {"type": "string"},
            "date_range": {"type": "string"},
        },
        "required": ["question"],
    },
}]

msg = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=1024,
    tools=tools,
    system="Answer only from auli_ask results and keep every citation.",
    messages=[{"role": "user",
               "content": "Summarise this week for the board in five lines."}],
)

claude mcp add auli --transport http https://mcp.auli.ai/mcp \
  --header "Authorization: Bearer $AULI_API_KEY"

# then, inside Claude Code:
#   > use auli to compare paid search CAC this month vs last

• Each connector has a dedicated dashboard with the metrics that platform reports natively.
• Metric selection and date range are saved per workspace, so the view you left is the view you return to.
• Any chart exports to PDF or CSV, and can be shared as a read-only link with people outside the workspace.
• Briefings can be delivered as a daily or weekly digest to the people who need the summary, not the tool.

Auli treats attribution as a first-class output. Every answer records the connectors it called, the accounts it read, the date range applied and the reconciliation rule used when sources disagree. Recommendations additionally state expected impact, a confidence level and the evidence chain behind them, so a reviewer can disagree with the reasoning rather than the tone.

• OAuth tokens are encrypted at rest and never exposed to the browser.
• Row-level security enforces workspace isolation in the database itself.
• Read-only scopes only — Auli cannot modify anything in a connected platform.
• Sign in with email and password, Google, Apple, or phone one-time code; social identities can be linked to an existing account later.
• Administrative actions are written to an immutable audit log with CSV export.

Each activated client account counts separately toward the connected accounts limit. Additional chat credits can be purchased at any time and are consumed only after the monthly allowance runs out. Full pricing lives on the pricing page.

Does Auli ever write to my platforms?

No. Every connector uses read-only scopes. Auli can read reporting data and cannot create, edit, pause or delete anything in Google Ads, Meta, LinkedIn, HubSpot or Shopify.

Where do the numbers in an answer come from?

Each figure carries its source system, the specific account or property, and the date range used. When two systems disagree, Auli reports the reconciled definition and the size of the gap rather than silently picking one.

Can one person work across several clients?

Yes. Switch workspaces from the top of the sidebar. Data never crosses a workspace boundary — row-level security is enforced in the database, not just in the interface.

What happens when I hit my message limit?

The analyst tells you before you run out and again at the limit. You can upgrade the plan or buy additional chat credits; credits are consumed only after the monthly allowance is used.