Every one is a real, working agent you can install today, not a demo. Pick your function, install it, and edit anything about it afterward.
Accounts Payable Manager, Accounts Receivable Collector
CRM Manager, Sales Pipeline Manager, Lead Qualification Specialist
Customer Support Resolver
SEO Operating System, LinkedIn Growth Manager, X (Twitter) Growth Manager
Order Intake Operator, Operations Report Operator
Web Researcher, Competitor Analyst
Email Manager, Meeting Notetaker
Executive Assistant, Business Analyst
Employee Onboarding Manager
Document Intelligence
Product Planner
Workflow Automation Engineer, Browser Operator
No lock-in on the model, no separate app for the channel, and no more access than the job needs for a teammate.
Pick a model per agent, per subagent, even per task. Switch anytime, nothing rebuilds.
An agent connects to WhatsApp, Slack, or Telegram and replies right there. No separate app for anyone to check.
Not a tool for just you. Invite a teammate by email, then control exactly which agents, providers, and tools they can see and touch.
A timer is enough for a simple recurring job. Real workflow work, the kind with fixed steps and branches, deserves a real workflow tool, so Oido treats n8n as a first-class citizen, not a bolt-on integration. The connection runs both ways.
An agent can trigger any n8n flow, even one with no webhook trigger of its own. The deterministic, many-step part of the job runs there, not squeezed into a prompt.
A flow calls an agent for the part it can't decide on its own: read a PDF or photo, check the wiki, rank results, judge an exception. The agent hands back an answer and the flow continues.
Install the community node, n8n-nodes-oido, in any n8n instance, yours or ours. Native n8n also runs inside Oido, so there's nothing separate to host.
n8n services →An agent that can act on its own is only worth deploying if you can prove what it is and is not allowed to do. Guardrails are rules your organization sets once, scanning everything an agent reads and writes, that no individual agent can loosen.
Block or flag on exact terms.
Regex for anything shaped, not just spelled.
Catches email, phone, SSN on the way in or out.
Broader classes of content, not single words.
Describe the policy in plain English, a model judges every message against it.
An organization-wide rule can only be tightened, never loosened, by a rule on a single agent.
Add an approval gate on top and the same principle carries to actions, not just text: a payment, a send, a delete pauses and asks a named human, every step logged.
How the rule builder works →Reads every incoming invoice, matches it to a PO, flags the two that don’t line up.
Output: a reconciled spreadsheet, 2 rows highlighted with the mismatch reason.
Scores an inbound lead against your ICP, drafts the reply, hands qualified leads to a rep.
Output: a CRM record updated with score and notes, a draft reply waiting for send.
Answers the repeat questions from your help docs, escalates anything it’s not sure of.
Output: a closed ticket with the answer sent, or an escalation note for a human.
Told to issue a refund, it stops, pings the named approver on Slack, and waits.
Output: an audit log entry, what it wanted to do, who approved it, when.
Agents read the wiki to answer accurately, and write to it as they learn. It's a knowledge graph, not a folder, so it connects itself as it grows instead of sitting there waiting to be organized.
New content is auto-linked into the graph on ingest, entity, concept, source, or synthesis. No manual tagging.
Keyword and vector search are fused, then reranked, so an agent gets the relevant page, not just the similar one.
What one agent works out gets written back. What it learns today, every agent starts from tomorrow, not lost in a chat.
A lint pass flags orphan pages, unresolved contradictions, and stale content, a running health check, not a one-time setup.
Org-wide vs agent-specific rules, with no override by design.
Describe a rule in plain English, get a working guardrail.
Control who on your team can edit the rules that govern every agent.
Watch input, model, and tool calls as a run happens, not after.
Redirect a running agent mid-task instead of only being able to stop it.
Every run, next to every rule it was checked against.
Build your first agent today at no cost. Scale to a whole team when you’re ready, your data stays yours throughout. One flat price covers everyone, no per-seat maths.
No. Bring OpenAI, Anthropic, Gemini, or your own model, and switch anytime. Oido is multi-model with no vendor lock-in, so you control your model costs.
No. Oido agents execute real work across your tools, send the email, update the record, run the report, open the ticket, not just answer questions.
One flat monthly price covers your whole team, not per seat, and not metered per task. The bill is the same whether you run a hundred jobs or a hundred thousand. You can start free and build your first agent at no cost.
Self-hosting is available on Enterprise. On other plans, Oido runs in a secure cloud where your data stays private and yours.
Gmail, Slack, HubSpot, Stripe, Google Drive, GitHub, Jira, your databases and REST APIs, plus any MCP server. Custom integrations are available on request.