Overview
A project is your workspace memory. An agent is your workspace intelligence. When you connect the two, the agent stops guessing and starts answering from the live workspace. New rows, edited tasks, updated dates. The agent sees them on the very next reply.
TL;DR: Open any custom AI agent, pick the Knowledge tab, click Add Project, and select a project from your workspace. The agent reads that project on every reply and stays in sync as the project changes. Remove the connection anytime from the same tab. This is the Memory to Intelligence arc of Workspace DNA, wired through the Knowledge tab.
Why connect a project
A static PDF answers from a snapshot. A live project answers from today.
- A connected sales pipeline lets the agent quote the right deal stage.
- A connected client roster lets the agent address people by name.
- A connected content calendar lets the agent know what already shipped.
The agent does not copy the project. It reads it. Edit a row in the morning and the agent reflects that change in the afternoon.
Memory feeding Intelligence
This is the first arc of Workspace DNA in real life. Projects (Memory) flow into the agent (Intelligence). Every reply pulls the latest workspace state.
Edit a row in the project, ask the agent a follow-up question, and the answer reflects the edit. No retraining. No re-upload. The agent re-reads the project on the next reply.
Add a project as a knowledge source
You set this up once per agent. Five clicks.
- Open the Agents tab in your workspace.
- Click the three dots next to the agent you want to wire up, then Edit agent.
- In the left sidebar, pick the Knowledge tab.
- Click Add Project and select one project from the picker. You can repeat this to add more.
- Click Update to save.
The agent now reads that project on every reply. A small backlink appears inside the project so anyone opening it can see which agents are using it as memory.
What the agent sees
Not every part of a project becomes knowledge. The reader is built for clarity, not noise.
| Included | Not included |
|---|---|
| Task text and block content | Comments and chat messages |
| Due dates | Reactions and timers |
| Assignees | Uploaded file attachments |
| Section structure | Custom field values |
If a custom field is mission-critical for the agent's answers (a deal stage, a status tag), restate it inside the task text so the reader picks it up. Files attached to a project belong on the agent's Media tab instead. See Agent Knowledge for the full source list.
Keep the connection fresh
The project stays live. You do not need to re-upload anything. A few small habits keep the agent crisp.
- Rename the project clearly. The title is the first thing the agent reads when it scans which source to consult.
- Keep one project per topic. A bloated project that mixes three projects in one will confuse the answer pattern.
- Review the agent's referenced sources after a few real chats. The agent shows which project it read for each reply.
Remove or swap a project
Connections are easy to undo. Nothing about your project is altered.
- Open the agent's Knowledge tab.
- Find the project in the list of sources.
- Click the row's remove control. The agent stops reading that project on the next reply.
Swap a project the same way. Remove the old one, click Add Project, pick the new one, hit Update. The agent picks up the new memory immediately.
Related guides
- Agent Knowledge & Memory. Every knowledge source type your agent can read.
- Custom AI Agents. Build the agent that consumes this memory.
- How Taskade Genesis Works: Workspace DNA. The bigger picture of Memory feeding Intelligence.
- Sweep Up Projects in One Move. Tidy your workspace before wiring up an agent.
- User Roles and Permissions. Who can connect projects to shared agents.