⚡ Agents · 6 production workers

The 6 Agents.

Agents are the action-taking half of OllaSuper. Each runs on a recurring schedule, uses real tools (ollagraph + filesystem + memory), and writes its output to a workspace you can read. Every action that could leave the platform is queued for your approval before it ships.

Looking for the chat-driven Experts? See the 80 Experts →

⚡ Agents vs ○ Experts

Agents run on schedules and take actions (queued for approval). Experts are chat-driven specialists with no schedule and no real-world side effects.

All 6 Sales 2 Marketing 2 Ops 1 Research 1

Sales

2 agents

⚡

Outbound SDR Agent

Researches prospects, drafts personalized cold outreach with company-specific hooks, queues every draft for your approval before send.

Every 30 minutes GPT-4o-mini Approval required

⚡

Pipeline Curator Agent

Reads your sales pipeline every 4 hours, flags genuinely at-risk deals, suggests the next specific action — not generic 'follow up'.

Every 4 hours GPT-4o-mini

Marketing

2 agents

⚡

SEO Auditor Agent

Real AEO + SEO audits every 6 hours, with specific code-level fixes — not generic 'improve your meta tags' advice.

Every 6 hours Tongyi DeepResearch

⚡

Content Publisher Agent

Researches deeply, then drafts AEO-optimized blog posts, social copy, and newsletters that are queued for your approval.

Continuous (manual trigger) GPT-4o Approval required

Ops

1 agent

⚡

Inbox Triage Agent

Every 15 minutes, classifies new inbound messages, drafts replies for routine ones, escalates the rest with full context.

Every 15 minutes GPT-4o-mini Approval required

Research

1 agent

⚡

Deep Research Agent

Multi-step research over the live web — structured briefs with citations. Powered by Tongyi DeepResearch, purpose-built for long horizons.

Continuous (manual trigger) Tongyi DeepResearch

⚡ How to use an Agent

Five steps from "schedule this" to a real action you sign off on.

Agents run on cron, call tools, draft real outputs (emails, PRs, audits, briefs) — then hold them at an approval gate so nothing ships without your nod.

STEP 1

Pick the Agent

Six production Agents auto-spawn at boot: Outbound SDR, Pipeline Curator, SEO Auditor, Content Publisher, Inbox Triage, Deep Research. Pick the one whose tools + role match your job.

STEP 2

Brief it like a teammate

Natural language. "Audit ollasuper.com for AEO regressions every Monday." The Agent's system prompt is already loaded with its workflow rules — you just hand it the target.

STEP 3

Watch it work

Agents call real tools: ollagraph scrape, web_fetch, file_write, memory_recall, MCP tools. The dashboard shows the tool-call trace so you can audit what it did and why.

STEP 4

Approve before it ships

Drafts queue in the approval surface — emails before send, blog posts before publish, PRs before merge. One click approves; one click sends back for a rewrite.

STEP 5

Schedule + hand off

Once you trust an Agent, turn on a cron schedule (every 15m / hourly / daily). Or have it hand off to another Agent — SDR finds leads, Outbound drafts emails, Inbox triages replies.

Schedules are off by default. A misconfigured cron loop on content-publisher (gpt-4o, smart tier) burned $27 in 4 days before we caught it. The 6 production Agents now ship queue-fed only — flip the schedule on per Agent when you actually want one. Use a per-day OpenRouter cap as a safety net.

⚡ Try an Agent

Talk to an Agent — they run real workflows

Every Agent has 4 starter prompts to try. Open the dashboard, pick an Agent, send your first prompt.

Open the dashboard →