Memory & Long-term Context

Files and meetings provide raw information. Memory is what your agent does with it.

Over time, your agents synthesize everything they learn into genuine understanding of your clients. They connect dots across documents, conversations, and interactions - building knowledge that grows more valuable the longer you use Verow.

How Agents Remember Over Time

Your agent is not just storing information. It is learning.

Every document you upload, every meeting you record, every conversation your client has with the agent - all of it contributes to a growing understanding. The agent notices patterns, remembers preferences, and builds context that spans months or years of client relationship.

This is fundamentally different from traditional file storage or CRM notes:

Traditional approach: You write down what seems important and hope you can find it later

Verow agents: The agent absorbs everything and surfaces what is relevant when you need it

After six months with a client, your agent knows their communication style, their concerns, their priorities, and the history of how projects have evolved. Ask it anything, and it draws on that accumulated understanding.

Memory Artifacts

As your agent processes information, it creates memory artifacts - distilled pieces of knowledge that persist across conversations.

What Memory Artifacts Capture

• Client preferences discovered over time
• Important decisions and their rationale
• Relationships between people and projects
• Patterns in client behavior and requests
• Key facts that come up repeatedly

How They Form

Memory artifacts emerge naturally. When your agent notices something significant - a stated preference, a decision made, a concern expressed - it creates a memory artifact. You do not need to tell it what to remember.

For example, if a client mentions in three separate meetings that they prefer conservative messaging, your agent forms a memory artifact about that preference. When you later ask it to draft content, it already knows to keep the tone measured.

Viewing Memory Artifacts

You can see what your agent has learned:

1. Open the agent
2. Navigate to Context > Memory
3. Browse the artifacts your agent has formed

This transparency helps you understand how your agent thinks and catch any misunderstandings early.

How Agents Build Understanding

Understanding develops in layers.

Layer 1: Raw Information

The foundation. Files, meeting transcripts, conversation logs, links. This is everything your agent has access to.

Layer 2: Extracted Knowledge

Your agent processes raw information and extracts specific facts, preferences, and relationships. "Client prefers email over Slack for approvals" is extracted knowledge.

Layer 3: Connected Understanding

This is where it gets powerful. Your agent connects knowledge across sources.

• A preference mentioned in a meeting links to decisions made in emails
• Project history from one campaign informs recommendations for the next
• A concern expressed early in the relationship explains behavior months later

Layer 4: Predictive Context

With enough history, your agent starts anticipating:

• What questions this client typically asks at this stage
• What concerns might arise based on past patterns
• What context will be most relevant for upcoming conversations

The Three Types of Context

Understanding how different context types work together helps you build more effective agents.

Files: Explicit Documents

Files are authoritative. When you upload a contract, the agent treats that as a source of truth. Files are:

• Uploaded intentionally
• Static until you update them
• Cited directly when relevant

Files answer questions like "What does the contract say about payment terms?" with direct references.

Meetings: Conversation History

Meetings capture how the relationship actually plays out. They include:

• Discussions that never made it to documents
• Tone and nuance of client interactions
• Informal preferences and concerns
• Real-time decisions and their context

Meetings answer questions like "Has the client ever expressed concern about our timeline?" by searching actual conversations.

Memory: Synthesized Knowledge

Memory is what your agent builds from files and meetings combined. It represents:

• Patterns noticed across multiple sources
• Preferences inferred from behavior
• Understanding that emerges over time
• Context that connects separate pieces of information

Memory answers questions like "What matters most to this client?" by drawing on everything the agent has learned.

Managing Agent Memory

You have control over what your agent remembers.

Reviewing Memory

Periodically review your agent's memory artifacts to ensure accuracy:

1. Open Context > Memory
2. Browse recent artifacts
3. Check that the agent's understanding aligns with reality

If an artifact seems incorrect, you can edit or remove it.

Correcting Misunderstandings

Agents can form incorrect memories, especially from ambiguous conversations. If you notice a wrong conclusion:

1. Find the problematic memory artifact
2. Click to edit or delete it
3. Add a correction if needed

For example, if the agent mistakenly believes a client prefers phone calls when they actually prefer email, correct that artifact. The agent updates its understanding accordingly.

Adding Manual Memories

Sometimes you know something the agent has not learned yet. Add manual memory artifacts for:

• Preferences learned outside of recorded interactions
• Important context from before you started using Verow
• Relationship dynamics you want the agent to understand

Navigate to Memory > Add Artifact and describe what the agent should remember.

Starting Fresh vs. Building on History

When to Build on History

Most of the time, you want your agents to accumulate knowledge. The longer the history, the better the understanding. Building on history works well when:

• Working with the same client over time
• Continuity matters for the relationship
• Context from past projects informs current work

When to Start Fresh

Sometimes a clean slate makes sense:

• Client relationship has fundamentally changed
• You are testing different approaches
• Old context is creating confusion
• You want to repurpose an agent for a different use

How to Reset Memory

To clear an agent's memory while keeping files and meeting transcripts:

1. Go to Context > Memory
2. Click Reset Memory
3. Confirm the reset

The agent keeps access to files and meetings but forgets its synthesized understanding. It rebuilds memory from scratch as new interactions occur.

Selective Reset

For targeted cleanup:

1. Browse memory artifacts
2. Select specific items to remove
3. Delete only what needs clearing

This preserves valuable understanding while correcting specific issues.

Memory and Team Collaboration

Memory belongs to the agent, not individual team members. When anyone on your team interacts with the agent:

• They benefit from accumulated memory
• Their interactions contribute to future memories
• They see the same memory artifacts

This means your newest team member can access the same client understanding as your longest-tenured account manager. The agent bridges knowledge gaps across your team.

Privacy Considerations

Memory artifacts are derived from your data, not copied from it. The agent's understanding might reference a conversation without storing the full transcript in memory.

All memory data receives the same security treatment as your files and meetings:

• Encrypted storage
• Team-only access
• No use in model training
• Full deletion available

When you delete an agent, all associated memory artifacts delete with it.