Appearance
Memories, AI Settings, Skills, and Runbooks
Junto gives you four ways to shape the agent's work. They are easy to mix up because each one changes what the agent knows or does, but they answer different questions:
- Memories: What should the agent remember about a person or a customer? Short facts saved from conversations.
- AI Settings: How should the agent behave everywhere? Standing rules and background about your MSP, set by an admin.
- Skills: How does our team handle this kind of request? Procedures the agent follows when a request matches.
- Runbooks: What work should Junto run, and when? Step-by-step automations that run from a ticket, a schedule, a webhook, or an email.
At a Glance
| Memories | AI Settings | Skills | Runbooks | |
|---|---|---|---|---|
| What it holds | A durable fact or preference about you, a customer company, or a contact | Agent language, Custom Instructions, and Organization Knowledge, for your MSP and per company | A procedure and the situations it covers | Executable steps that call your integration tools |
| Example | "Acme's front desk printer is on the guest VLAN." | "Always be formal in communications." | "Printer offline: check the spooler, clear stuck jobs, send a test page." | "Reset a user's MFA and post the result to the ticket." |
| Who creates it | The agent, when you ask it to remember something or it learns a lasting fact. You can also add one yourself. | Admins and Owners | Admins and Owners | Your team, in the runbook editor or by asking Explore |
| When it's used | Recalled as context in later Explore conversations | Every time the agent works | Only when a request matches the skill | When triggered: a matched ticket, Run now, a schedule, an incoming webhook, or a mailbox trigger |
| Where it applies | Explore in the web app, Slack, and Teams | Ticket triage, ticket chat, and Explore; Agent language covers everything the agent writes | Explore and connected AI clients | Runbook runs, on a ticket or for a company |
| Does it do anything on its own? | No. It is context only. | No. It guides behavior. | No. It guides a conversation. | Yes. It runs, within your tool policies. |
None of the four grants new permissions. Every action the agent takes still goes through your Tool Policies, with the same Minimum role and Approval settings as any other request.
Which One Should I Use?
| If you want the agent to... | Use |
|---|---|
| Know that a contact prefers phone over email, or that a customer's server room is off-site | A company or contact memory |
| Answer you in bullet points, or remember which clients you're responsible for | A private memory |
| Write in Spanish, keep a formal tone, or escalate when uncertain, on every ticket | Custom Instructions in AI Settings |
| Know your business hours, your timezone, or which RMM you standardize on | Organization Knowledge in AI Settings |
| Follow a different rule for one customer only | Company-level AI settings on that company's AI tab |
| Work a "new hire setup" or "printer offline" request your team's way | A skill |
| Do the work itself, on a matched ticket or on a schedule, with the steps fixed in advance | A runbook |
A good rule of thumb: a memory is a fact, a setting is a standing rule, a skill is a procedure the agent follows when it fits, and a runbook is a procedure Junto runs.
Memories
Memories let the agent carry what it learns from one Explore conversation into the next. Each memory is a short note with a title and a body, saved in one of three places:
| Where it's saved | Who sees it | What belongs there |
|---|---|---|
| Private (about you) | Only you | How you like answers, what you're responsible for, your standing preferences |
| Company | Every technician who works with that customer | Environment quirks, standing procedures, escalation contacts |
| Contact | Every technician who works with that contact's company | A person's role, preferences, or access notes |
How memories are saved
Tell the agent what to keep, for example "Remember that Acme's after-hours contact is the office manager, not the owner" or "From now on, give me the short version first." The agent may also save a durable fact it learns about a customer that none of your systems record. It tells you in one sentence whenever it saves, changes, or removes a memory, and which folder it used. Before adding a memory, it checks for an existing one on the same topic and updates that instead of creating a duplicate.
- Private memories save without a prompt.
- Saving or changing a company or contact memory waits for your approval. Because the whole team sees them, the approval card shows the folder, the title, and the full text before anything is saved. If you decline, the agent won't try to save it again. Removing a memory doesn't need approval.
The agent doesn't save one-off ticket details, passwords or other secrets, or anything your PSA or documentation already records. Shared memories hold operational facts only: the agent never saves personnel, HR, legal, billing-dispute, or performance matters, or opinions about people, to a company or contact. If you explicitly ask it to remember something like that, it goes in your private memories only.
How memories are used
At the start of each Explore conversation, the agent loads your private memories. As soon as the conversation turns to a specific customer company or contact, it looks up that company's or contact's memories before it answers or acts.
Memories are treated as context, never as instructions. A memory that says "Acme's owner approves all purchases" informs the agent; it doesn't change what the agent is allowed to do. To change behavior, use AI Settings or a skill.
Memories are used in Explore in the web app, Slack, and Teams. They are not used in Slack channels shared with another organization, and customers never see them or have their own.
Viewing and editing memories
- Your private memories: open Preferences from the user menu. The Memory card lists what Junto remembers about you. You can add, edit, and delete them.
- Company and contact memories: open the company and select the Memory tab. It shows the company's shared memories and, under About contacts, the memories about each of its contacts. Anyone on your staff who can see the company can read them; Operators and above can add, edit, and delete them.
Each folder holds up to 50 memories. A title can be up to 120 characters and a memory up to 4,000 characters. Every change is recorded in your audit log; for private memories, the title is left out so the log doesn't reveal what the note is about.
Memories vs. chat history
Memories are separate from your conversation history. Reopening an Explore session continues that conversation; starting a new one begins fresh, apart from your memories and settings. See Explore sessions.
AI Settings
AI Settings are the standing rules for your whole organization. Admins and Owners manage them under Settings > AI Settings > Configuration:
- Agent language: the language the agent writes in everywhere.
- Custom Instructions: how the agent should behave, such as "Always be formal in communications" or "Prioritize security-related tickets."
- Organization Knowledge: background about your MSP, such as your business hours, timezone, cloud provider, and critical systems.
Each company also has an AI tab with company-level settings that add to (not replace) the organization settings for that customer's tickets.
Unlike memories and skills, AI Settings apply whenever the agent works: in ticket triage, ticket chat, and Explore. Agent language also covers customer replies, runbook output, and summaries. Changes take effect on the next agent interaction. See the AI Settings FAQ for details.
Settings or memory?
If a rule should apply to every ticket for a customer, put it in the company's AI tab, where an admin owns it. If it's a fact the team picked up along the way, a company memory is the lighter-weight choice.
Skills
A skill is a procedure your team writes once: a name, a description of when the agent should use it, and the instructions to follow. The agent reads the skill only when a request in Explore or a connected AI client matches it, so you can keep many detailed procedures without crowding every conversation.
Use a skill when the how matters, for example the order your team checks things in, the questions to ask first, or what to record in the ticket. Admins and Owners create skills under Settings > AI Settings > Skills. See Skills.
Runbooks
A runbook is work Junto runs, not guidance it reads. Each step can call your integration tools, and a runbook can start from a ticket that triage matches to it, from Run now, on a schedule, from an incoming webhook, or from a mailbox trigger. New runbooks run supervised, with a technician approving each launch, and can move to auto-run once you trust them.
Use a runbook when you want the outcome to happen the same way every time, with a record of each run. See What Are Runbooks?.
Skill or runbook?
If a technician is in the conversation and the agent should follow your process while it helps, write a skill. If you want Junto to carry out the steps itself, on a matched ticket or a schedule, build a runbook.
Putting It Together
Here's how all four might shape one printer ticket from Acme:
- Organization Knowledge tells the agent your business hours, so it doesn't promise an on-site visit at 9 p.m.
- Custom Instructions keep the reply formal.
- A company memory reminds it that Acme's front desk printer sits on the guest VLAN.
- A "Printer offline" skill gives a technician in Explore your team's troubleshooting order.
- A "Restart print spooler" runbook matches the ticket and, once approved, restarts the spooler and posts the result.
FAQ
Can customers see memories, settings, or skills?
No. Memories and skills are never available to customer accounts, and AI Settings are only visible to Admins and Owners. Customers see only the agent's replies.
Can a memory override my tool policies or AI Settings?
No. Memories are context, not instructions, and none of the four can grant permissions. Tool policies decide what runs and who approves it.
The agent saved something wrong. How do I fix it?
Ask the agent to update or forget it, or edit it directly: your private memories under Preferences > Memory, and company or contact memories on the company's Memory tab (Operator or above).
Why didn't the agent use my skill?
Skills apply only when a request matches the skill's When should Junto use this? text, and only in Explore and connected AI clients, not in ticket chat or runbook runs. See Skills.