Agents Overview
Agents that act. Not chat.
Ariftly ships vertical AI agents — agents that already understand their domain and are ready to work the moment you connect your tools. There's an agent for everyone: compliance leads, sales reps, founders, engineers, and operators.
Unlike workflow builders where you wire up logic yourself, Ariftly agents arrive with domain knowledge, tool integrations, and decision-making built in. You describe what you want; the agent does the work and asks for approval before anything is sent or changed.
Available Agents
| Agent | Who It's For | Status |
|---|---|---|
| AI Readiness Agent | Compliance leads, legal teams, anyone answering EU AI Act or procurement questionnaires | ✅ Live |
| Sales Agent | Sales reps and GTM teams — lead discovery, enrichment, AI-drafted outreach | ✅ Early Access |
| Finance Agent | Finance and ops teams — invoice reconciliation, spend anomaly detection | 🔜 Coming soon |
| DevOps Agent | Engineers and SREs — incident triage, deployment risk scoring | 🔜 Coming soon |
| Founder Agent | Founders and operators — competitive intelligence, board pack generation | 🔜 Coming soon |
How Agents Work
Every agent follows the same lifecycle:
Task triggered (manual / schedule / channel / API / MCP)
│
▼
Agent receives TaskTriggerRequest
(credentials, tools, knowledge, policies, AI gateway)
│
▼
Agent executes domain logic
(AI reasoning + tool calls via proxy)
│
▼
Agent emits Events back to Core
(progress, artifacts, approval requests)
│
▼
You review findings in dashboard or Slack
│
▼
You approve → action executed
What Every Agent Gets
Every deployed agent automatically has access to:
| Capability | What it is |
|---|---|
| AI Gateway | Multi-provider AI routing with your own API key (BYOK) |
| Tool Proxy | Pre-built connectors to your connected SaaS tools — scoped, proxy-only, never raw secrets |
| Knowledge | Tenant-scoped document store for retrieval-augmented generation |
| Memory | Long-lived key-value and time-series state across tasks |
| Approvals | Human-in-the-loop gates for any external action |
| Skills | Extend agent behavior by describing what you want in plain English |
| Telemetry | OpenTelemetry traces on every Core ↔ Agent ↔ Tool call |
Agent Colors
Each agent has a dedicated visual identity across the dashboard and docs:
| Agent | Color | Token |
|---|---|---|
| AI Readiness Agent | Indigo / violet #818CF8 | agent-ai |
| Sales Agent | Brand orange #F97316 | agent-sales |
| Coming soon agents | Slate grey | agent-coming |
Channels
Every agent can receive tasks and send results through any channel — all normalized to the same message format:
- Dashboard — web UI task runner and approval inbox
- Slack — approve drafts, receive findings, trigger tasks with
/ariftly - Email — forward questionnaires directly to the AI Readiness Agent
- API —
POST /v1/tasksfor programmatic task dispatch - Schedule — cron-based recurring tasks
- MCP — Claude Desktop, ChatGPT, Cursor can invoke agents directly
Adding a Custom Agent
If the built-in agents don't cover your domain, you can build and deploy your own agent using the Ariftly SDK and the Remote Agent Protocol (RAP v1). A new agent is ~200 lines of domain code on top of the SDK scaffold.