# Agent Knowledge & Memory

***

## Table of Contents

* [Overview](#overview)
* [Explore the Knowledge Tab](#explore-the-knowledge-tab)
* [Train Your Agent](#train-your-agent)
* [Supported Knowledge Sources](#supported-knowledge-sources)
* [Training Methods](#training-methods)
  * [Upload Documents](#upload-documents)
  * [Use Web Resources](#use-web-resources)
  * [Create Sources Manually](#create-sources-manually)
  * [Dynamic Knowledge (Projects)](#dynamic-knowledge-projects)
* [Bulk Training](#bulk-training)
* [Look Up References](#look-up-references)
* [Knowledge Backlinks](#knowledge-backlinks)
* [What Projects Expose to Agents](#what-projects-expose-to-agents)
* [Knowledge Best Practices](#knowledge-best-practices)
* [What's Next](#whats-next)

***

## Overview

Agent knowledge is the information your custom AI agents use to provide **contextual, accurate responses**. It combines pre-trained AI capabilities with **your own data** — documents, URLs, projects, and media — to create agents that truly understand your business.

> **Workspace DNA connection:** Knowledge is what transforms a generic AI into a specialist. The more knowledge you add, the richer your [Workspace DNA](/taskade-genesis/genesis/how-genesis-works.md) becomes, and the smarter all your Genesis apps get.

***

## Explore the Knowledge Tab

| Step | Action                                      |
| ---- | ------------------------------------------- |
| 1    | Go to the **Agents** tab                    |
| 2    | Click the three dots (⋯) next to your agent |
| 3    | Select **Edit agent**                       |
| 4    | Open the **Knowledge** tab                  |

The Knowledge tab shows all sources your agent has been trained on, with options to add, remove, and manage sources.

***

## Train Your Agent

### Supported Knowledge Sources

| Source Type          | Formats / Types                                     | How It's Used                                   |
| -------------------- | --------------------------------------------------- | ----------------------------------------------- |
| **Documents**        | PDF, CSV, TXT, DOCX, MD, PPTX, XLSX, EPUB           | Full text extraction and indexing               |
| **Web URLs**         | Any web page, blog post, documentation page         | Content scraped and indexed                     |
| **YouTube videos**   | Video URLs                                          | Transcript extracted and indexed                |
| **Social content**   | Tweets, Reddit posts, news articles                 | Content extracted and indexed                   |
| **Cloud files**      | Google Drive, Dropbox, Box, OneDrive                | File content pulled and indexed                 |
| **Taskade projects** | Any project in your workspace                       | Live, dynamic knowledge (updates automatically) |
| **Manual sources**   | Markdown or plain text created in the Knowledge tab | Direct knowledge injection                      |

***

## Training Methods

### Upload Documents

| Step | Action                                             |
| ---- | -------------------------------------------------- |
| 1    | Open the Knowledge tab                             |
| 2    | Drag & drop files directly, or click **Add Media** |
| 3    | Select files from your device                      |
| 4    | Files are processed and indexed automatically      |

**Supported document formats:**

| Format         | Best For                                        |
| -------------- | ----------------------------------------------- |
| **PDF**        | Brand guidelines, reports, manuals, legal docs  |
| **CSV / XLSX** | Customer data, product catalogs, pricing tables |
| **DOCX**       | Procedures, policies, templates                 |
| **TXT / MD**   | Plain text knowledge, FAQ answers, notes        |
| **PPTX**       | Presentations, training materials               |
| **EPUB**       | Books, long-form content                        |

### Use Web Resources

| Step | Action                                                 |
| ---- | ------------------------------------------------------ |
| 1    | Click **Add Media** in the Knowledge tab               |
| 2    | Paste a URL (web page, YouTube video, cloud file link) |
| 3    | Content is scraped/transcribed and indexed             |

| Resource Type                      | What Gets Indexed         |
| ---------------------------------- | ------------------------- |
| **Web pages**                      | Full page text content    |
| **YouTube videos**                 | Full video transcript     |
| **Google Drive files**             | Document text content     |
| **Dropbox / Box / OneDrive files** | Document text content     |
| **Blog posts**                     | Article text and metadata |
| **Tweets**                         | Tweet text and thread     |
| **Reddit posts**                   | Post and top comments     |
| **News articles**                  | Full article text         |

### Create Sources Manually

| Step | Action                                                   |
| ---- | -------------------------------------------------------- |
| 1    | Open the Knowledge tab                                   |
| 2    | Click **"Create Source"**                                |
| 3    | Write or paste Markdown/plain text directly              |
| 4    | Save — the source is immediately available to your agent |

> **Tip:** Use manual sources for FAQ answers, custom instructions, edge case handling, or information not available in any file.

### Dynamic Knowledge (Projects)

Connect Taskade projects as **living knowledge sources** that update automatically:

| Step | Action                                                            |
| ---- | ----------------------------------------------------------------- |
| 1    | Open the Knowledge tab                                            |
| 2    | Click **Add Media** → **Add from Workspace**                      |
| 3    | Select one or more projects                                       |
| 4    | Projects become live knowledge — agent always has the latest data |

> **Dynamic = real-time.** When you update a project, the agent's knowledge updates automatically. No re-upload needed.

***

## Bulk Training

Upload multiple sources simultaneously:

| Method                    | How                                                    |
| ------------------------- | ------------------------------------------------------ |
| **Multi-file upload**     | Select multiple files in the file picker               |
| **Multi-project connect** | Select multiple projects from your workspace           |
| **Multi-URL paste**       | Add multiple links at once                             |
| **Drag & drop batch**     | Drag a folder or multiple files into the Knowledge tab |

***

## Look Up References

When your agent responds, it shows which sources it referenced:

| Feature                | Detail                                                   |
| ---------------------- | -------------------------------------------------------- |
| **Source citations**   | Chat displays which knowledge sources were used          |
| **Preview capability** | Click a citation to preview the source content           |
| **Transparency**       | Users can verify agent responses against source material |

***

## Knowledge Backlinks

Projects used as agent knowledge show **backlinks** indicating which agents reference them:

| Feature                      | Detail                                                           |
| ---------------------------- | ---------------------------------------------------------------- |
| **Backlink indicator**       | Projects display a badge showing connected agents                |
| **Bi-directional awareness** | See which agents use a project, and which projects an agent uses |
| **Impact visibility**        | Know which agents are affected when you edit a project           |

> **Tip:** Backlinks help you maintain knowledge integrity. Before editing a project, check its backlinks to understand which agents depend on it.

***

## What Projects Expose to Agents

When a project is connected as knowledge, agents can access:

| Accessible                                    | Not Accessible                               |
| --------------------------------------------- | -------------------------------------------- |
| Text content (tasks, notes, descriptions)     | Uploaded file attachments within the project |
| Block structure (headings, lists, checklists) | Comments on tasks                            |
| Due dates                                     | Embedded links within task descriptions      |
| Assignee information                          | Reactions (emojis)                           |
| Timer data                                    | Chat messages in the project                 |
| Custom field values                           |                                              |
| Labels and tags                               |                                              |

***

## Knowledge Best Practices

| Practice                           | Why                                           | Example                                               |
| ---------------------------------- | --------------------------------------------- | ----------------------------------------------------- |
| **Start with your most-used docs** | Highest ROI knowledge sources                 | FAQ doc, product guide, pricing sheet                 |
| **Use projects for dynamic data**  | Always up-to-date, no re-uploading            | Customer list project, inventory project              |
| **Create manual sources for gaps** | Fill knowledge holes quickly                  | Edge case handling, tone guidelines                   |
| **Organize by topic**              | Helps agent find relevant info faster         | Separate "Product" and "Pricing" sources              |
| **Update regularly**               | Stale knowledge = wrong answers               | Quarterly review of all knowledge sources             |
| **Test with real questions**       | Verify agent uses knowledge correctly         | Ask questions you know the answers to                 |
| **Use backlinks to audit**         | Understand cross-dependencies                 | Check before editing shared projects                  |
| **Upload diverse formats**         | Different formats capture different knowledge | PDFs for detail, CSVs for data, URLs for current info |

***

## What's Next

| Guide                                                                     | What You'll Learn                            |
| ------------------------------------------------------------------------- | -------------------------------------------- |
| [Tools for AI Agents](/ai-features/ai-features/agent-tools.md)            | Add action capabilities to your agents       |
| [Custom AI Agents](/ai-features/ai-features/ai-agents-getting-started.md) | Build and configure agents from scratch      |
| [Multi-Agents](/ai-features/ai-features/multi-agents.md)                  | Create teams of agents with shared knowledge |
| [Projects & Databases](/workspaces/workspaces/projects-databases.md)      | Structure data for optimal agent training    |
| [Adding Genesis Context](/taskade-genesis/genesis/adding-context.md)      | Upload context for Genesis app creation      |


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.taskade.com/ai-features/ai-features/agent-knowledge.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.