Create and Configure Agents
Set up an AI agent that runs your workflows automatically. AI operates on your behalf — you supervise, review, and approve.
What is an agent?
An agent is an AI that executes workflows on a schedule without you having to trigger each run manually. It reads context from your Knowledge Graph, takes actions through your connected apps, and checks back with you before doing anything irreversible. You stay in control — the agent handles the repetitive work.
Create an agent — 5-step wizard
Open the AgentLed web app and navigate to Agents → New Agent. The wizard walks you through five steps.
Type
Choose what kind of work this agent will do. Options include deal sourcing, content publishing, lead enrichment, ops monitoring, and custom. Selecting a type pre-fills sensible defaults for the next steps — you can always change them.
Instructions
Write plain-language instructions that tell the agent what to do, what to prioritize, and what to avoid. Think of it as an onboarding document for a new hire. Be specific about your goals and the output you expect.
// Example instructions
Find high-intent fintech leads in Europe. Score each lead
against our ICP (Series A+, 50–500 employees, DACH or UK).
Only add leads with a score above 70. Flag anything
uncertain for my review before adding to the pipeline.
Tools
Select which apps and integrations the agent can use — LinkedIn, HubSpot, Hunter, Slack, Notion, and 100+ others. The agent will only have access to tools you explicitly grant here. You can add or remove tools at any time after creation.
Workflows
Assign one or more existing workflows to the agent. The agent will run these workflows as part of its work cycle. You can create new workflows first from the Workflows section, or attach them after the agent is created.
Schedule
Set how often the agent runs. Choose a preset (every hour, daily at 8am, weekly on Monday) or enter a custom cron expression. You can also leave the schedule off and trigger runs manually or via the API.
Connect workflows
Workflows are the building blocks an agent executes. To connect a workflow to an existing agent:
- 1.Open the agent from Agents and go to the Workflows tab.
- 2.Click Add Workflow and select from your existing workflows, or create a new one inline.
- 3.Set the run order if multiple workflows are attached — the agent executes them in sequence.
- 4.Save. The agent will use the updated workflow list on its next scheduled run.
Each workflow runs as its own execution. Results from one workflow are stored in the Knowledge Graph and are available to the next workflow in the sequence.
Schedule and run autonomously
Once the schedule is set, the agent runs without any intervention needed. Each run:
- —Reads relevant context from the Knowledge Graph (what ran before, what was scored, what was flagged)
- —Executes each attached workflow in order
- —Writes results back to the Knowledge Graph so the next run builds on this one
- —Pauses at any human-in-the-loop step and notifies you before continuing
You can trigger a manual run at any time from the agent detail page by clicking Run Now. Useful for testing before enabling the schedule.
Chat and approve
Agents do not act without oversight. When a workflow hits a human-gate step — sending an email, writing to a CRM, publishing content — the agent pauses and sends you a notification.
From the agent chat panel you can:
- —Approve — the agent continues and takes the action
- —Reject — the step is skipped, the agent moves on
- —Edit and approve — modify the output before it goes out
- —Ask a question — chat with the agent about why it made a decision
The agent never publishes, sends, or writes to external systems without passing through a human gate — unless you explicitly configure it as fully automated.
Real example: Inovexus deal sourcing agent
Inovexus is a VC firm that uses AgentLed to source deals automatically. Their agent runs every morning and handles the full top-of-funnel pipeline — no analyst time required for sourcing.
Inovexus — Deal Sourcing Agent
The accuracy improvement comes from the Knowledge Graph — each approved and rejected lead teaches the agent what a good fit actually looks like for Inovexus specifically.
Next steps
- Human-in-the-Loop — Configure approval gates in your workflows
- Persistent Memory — How the Knowledge Graph improves accuracy over time
- Proactive Agents — Trigger agents based on events, not just schedules
- Knowledge Graph Overview — Understand what context your agent can access
