Agents help — the simple version

This page explains how agents work on iFusionSoft. Think of agents as helpful robots that check your apps, follow rules you wrote, and do small jobs on a schedule—so you don’t have to click the same things every day.

What are agents?

A chat is like asking one question and getting one answer. An agent is different: it stays connected to your tools (email, calendar, Slack, LinkedIn, and more), remembers its rules, and can run again tomorrow without you asking.

Words we use on this page

Agent is the real product term. Robot on this page just means the automatic part—not a separate thing you install.

Word in help	What it really means
Agent	The real thing: a configured worker (email analyst, LinkedIn poster, timeline builder, etc.) with rules, connections, schedule, and reports.
Robot (casual)	The automatic part — runs without you clicking each time, gathers data, merges timelines.
You / your team	The judgment part — review chronology entries, approve bridges, fix YAML, sign in with OAuth.

Example: “Every morning, read my inbox, find urgent customer emails, and save a short summary.” That’s agent work—not a single chat message.

Kid-simple rule: If it should happen again and again, use an agent. If you only need an answer right now, chat is fine.

Three lanes (pick your path)

On the Agents tab you’ll see main ways to work:

Hub lane

Recurring jobs on a schedule. Connect apps once, add instances, turn on bridges. Best for mail, Slack, CRM, LinkedIn, workplace automation.

Research Library + SharePoint

Model Pile debate/critique/vision runs linked to a SharePoint project folder—sources, exports, and consensus for your team.

Desktop / AgentTasker

One-off tasks queued from the desktop app—like “do this step for me once” with a human in the loop.

Marketing tools

Paste-and-go helpers: decks, briefs, campaign copy. Great for creative work, not for 24/7 monitoring.

Use I want to… on the Agents tab if you’re not sure which lane fits—it now includes a Research Library + SharePoint group for project document workflows.

First-time setup (5 steps)

Open Automation Hub from the Agents tab.
Go to Connections and sign in to the app you need (Microsoft, Google, Slack, etc.).
Go to Add instance (or install a Marketplace pack) to create your agent’s folder.
Open the instance editor, set your rules in YAML (or use AI Generate to help write them), then save.
Press Run once to test. Check Cycle Reports to see what happened.

Sign-in matters: Many agents need OAuth (the “Sign in with Microsoft/Google” flow). If an instance says “needs sign-in,” complete that before expecting results.

Automation Hub basics

The Hub is your control room. Main tabs:

Overview — health at a glance (what’s OK, what needs attention).
Timeline — a live feed of what agents and bridges just did (see below).
Chronologies — merged, dated event logs you review and export (not the same as Timeline).
Instances — each running copy of an agent (one per mailbox, brand, or team).
Connections — shared sign-ins you can reuse across instances.
Bridges — wires between agents (“when A finishes, tell B”).
Marketplace — install packs (e.g. workplace automation) in one click.

Timeline vs Chronologies — what’s the difference?

These two Hub tabs sound similar, but they do different jobs. Easy way to remember:

Timeline tab

“What happened lately?” A scrolling news feed of agent runs and bridge deliveries—newest first. Approve dry-run bridges here too.

Chronologies tab

“Put the story in order.” A merged dated log built from mail, files, and bridges—then your team reviews, links related events, and exports.

Short version: Timeline = activity log. Chronologies = curated timeline you build and check before sharing.

Timeline tab (activity feed)

Open Automation Hub → Timeline. You’ll see a unified list of:

Agent runs — “email analyst finished,” “LinkedIn poster ran,” and so on.
Bridge deliveries — especially dry-runs waiting for Approve live.

Think of it like a security camera replay for your robots: what ran, when, and whether it succeeded. It’s similar to the Activity trail on the main Agents tab, but inside the Hub with bridge approvals built in.

Chronologies (merged timelines)

Chronologies are for when you need one clean, date-ordered list of events pulled from many sources—not just “what ran five minutes ago.”

Mail, documents, and bridge feeds can all feed a Timeline builder instance (agent type: case-chronology-builder). Automation merges them into a master log. Then you:

Review each entry (confirmed, disputed, needs research, etc.).
Link related events (“this email goes with that meeting”).
Export when ready (e.g. Word doc for a case, ops review, or project recap).

Not only for lawyers

The tab works for general timelines by default. Legal teams can use the same tools with matter-specific labels; operations and project teams can track incidents, rollouts, or milestones the same way.

Getting started: Add a Timeline builder instance from Instances, or install a marketplace pack (e.g. legal chronology). Follow the onboarding runbook on the Chronologies tab: connect → bridges → merge → review → export.

Don’t mix these up: A LinkedIn instance named “Legal Chronology” is not a chronology timeline. You need a case-chronology-builder:… instance for the Chronologies tab.

Why review chronology entries?

Merge is what robots do: collect events from mail, documents, and bridges. Review is what people do: decide what you’re willing to stand behind before you export or share.

Automation is great at gathering—but not perfect. Sources can disagree, dates can be fuzzy, and the same event can show up twice. Review is how you turn “here’s what we scraped” into “here’s our official timeline.”

What robots miss

Conflicts — two emails describe the same day differently.
Duplicates and noise — invite, reminder, and follow-up look like three events.
Bad or missing dates — “last Tuesday” or forwarded threads need a human call.
Missing context — you know a message was routine, sarcastic, or already resolved.

What review captures

On each entry you can set:

Verified / Confirmed — checked; belongs on the timeline.
Questioned / Disputed — not sure, or conflicts with another source.
Needs follow-up — flag for more research.
Links — “this email supports / contradicts that meeting.”
Known as of — when your team actually knew about it (important for legal and incident timelines).

Use the Quick note dropdown under the review box to insert a preset sentence for the status you picked (Verified, Disputed, etc.), then edit if needed and click Save review.

One-line rule: Timeline tab = what ran recently. Chronologies = the story you’ll export after humans verify it.

Why it matters

Legal — counsel needs a defensible chronology, not an unvetted inbox dump.
Operations — postmortems need verified facts and “what we knew when.”
Projects — stakeholders need a clean milestone story, not mail clutter.
General — any shared timeline should reflect human judgment before you send it out.

Connections (sign in once)

A connection is like a shared key to an app. You sign in once, then many agent instances can use it. That’s easier than signing in separately for every mailbox or team.

On an instance row, use Assign vault to pick a shared connection, or leave it on Profile OAuth if that instance should sign in on its own.

Instances (one agent, many copies)

An instance is one copy of an agent—like one Lego set built from the same kit. Example: one LinkedIn poster for “Acme Corp” and another for “AlgoVision.”

Display name

The display name is the friendly label you see in lists (e.g. “Acme mailbox”). The slug is the folder name on disk (e.g. acme-mailbox)—lowercase, hyphens only.

When editing an instance, change the display name in the toolbar above the editor, or edit profile.display_name in YAML and save.

Schedule vs run once

Run (lightning bolt) — do the job right now.
Play / Stop — turn scheduled runs on or off.

Bridges (agents talking to each other)

A bridge listens for an event from one agent and triggers another. Example: “When email analyst finishes, post a summary to Slack.”

Bridges often start in dry-run mode: they show you what would happen without sending anything live. Use Approve in the Activity trail when you’re ready for real delivery.

Examples you can add from Marketplace or I want to…:

SharePoint file → Slack — notify the team when a new file appears.
Research library source → debate prep — when a doc lands in your project sources folder, log a prep note for debate/critique workflows.
Mail digest → chronology — feed the Timeline builder for reviewed exports.

Cycle reports

After each run, agents can write an HTML report—a small webpage that says what they did. Open Cycle Reports in the Hub (or the report button on an instance) to read them.

If reports are empty, run the agent once manually and check for sign-in or config errors on the Overview tab.

Activity trail

The Activity trail on the main Agents tab is like the Hub’s Timeline tab: what ran, what was drafted, and what is waiting for you to Approve.

Use Preview to see a draft, Replay to run again, and Approve when you trust it to go live. For bridge approvals inside the Hub, you can also use Timeline there.

The I want to… picker on the same Agents tab lists common paths—including Hub setup, Research Library + SharePoint, AgentTasker one-shots, and Marketing tools—with a short “why” and starter steps for each.

Chronologies are different—they’re merged, reviewable event timelines, not the live activity feed. See Timeline vs Chronologies above.

Workplace automation agents

Five agents for everyday office work:

Meeting → actions — turn meetings into tasks and follow-ups.
Contract / vendor inbox — triage vendor and contract email.
RFP / questionnaire — draft answers from your knowledge.
Pipeline hygiene — keep CRM deals tidy and flagged.
Expense / receipt — pull receipt info from mail.

Install the Workplace automation (top 5) pack in Marketplace, or add instances one at a time from Add instance.

When something breaks

“Needs sign-in”

Copy the OAuth URL from the instance row (key icon) and complete sign-in in your browser.

“Nothing in reports”

Run once manually. Open the editor and check YAML for typos. Make sure scheduling isn’t paused.

“Display name didn’t stick”

Use the display name field in the editor toolbar and click save, or edit YAML and save—then refresh the Hub.

Still stuck?

Email support@ifusionsoft.com with the instance name and what you expected to happen.

Project Research Library + cloud folder

In Model Pile, the Research Library sidebar saves your compare, critique, debate, and vision runs. You can link that library to OneDrive/SharePoint (sharepoint-organizer) or Google Drive (drive-organizer) so documents and reports live in a real project folder—not just on this browser.

What you get

Project folders — each project gets sources, exports, and consensus subfolders in your cloud drive.
Save sessions — click the cloud export icon on a Research Library session to export a markdown report.
Reuse documents — browse project files in the sidebar and open them directly in Debate or Critique (no re-hunting in email).
Docs Office round-trip — open a Word or deck file from the project folder; Ctrl+S saves back to the same cloud file.
Agent knowledge — drop source PDFs in sources; the organizer agent can emit bridge events when new files appear.

Setup (two minutes)

In Agents, add a SharePoint organizer or Drive organizer instance and sign in.
In the organizer config, enable research_library.enabled: true (on by default in example profiles).
In Model Pile, open Research Library → pick provider → pick agent → project name → Link folder.

Find it in I want to…

On the Agents tab, open I want to… and look for these chips:

Connect SharePoint for project research folders — Hub setup (SharePoint organizer + YAML).
Link my Research Library to Google Drive — Drive organizer + Research Library panel.
Link my Research Library to a SharePoint / OneDrive project — jumps to Compare and the Research Library panel.
Debate or critique a doc from project cloud sources — use project files without re-uploading.
Save a debate / critique session to project cloud folder — export a session report to exports.

The same Microsoft connection can live in the Agents vault (connection_id) so multiple organizer instances share one sign-in.

Docs Office AI + project cloud folder

Docs Office AI (Word) and Docs Office Presentation use the same project cloud folder as Model Pile Research Library (OneDrive/SharePoint or Google Drive). Look for the cloud button beside My Documents / My Presentations.

What you can do

Open from cloud — browse project sources or exports and load a .docx, .html, .pptx, or .html deck into the editor.
Round-trip save — files opened from the project folder register for round-trip; Ctrl+S saves back to the same cloud file (plus local My Documents).
Save copy to exports — use Save copy to exports in the cloud modal to write a new file into exports without overwriting the source.
One link everywhere — the project folder you link in Doc Office is the same config used in Model Pile Research Library.

Setup

Connect a SharePoint organizer or Drive organizer agent in Automation Hub.
In Doc Office, click the cloud toolbar button → pick provider → agent → project name → Link folder.
Use Refresh files to browse and click a file to open (round-trip is enabled automatically).

Note: Round-trip runs on Ctrl+S / Save to My Documents. Use Save copy to exports when you want an additional copy in the team folder.

Word list

Agent: A program that follows rules and can run on a schedule.
Instance: One running copy of an agent (one profile folder).
Connection / vault: Shared sign-in to an app, reused by instances.
Bridge: A rule that connects two agents (“when A happens, do B”).
YAML: The config file format agents use—like a settings list with indentation.
OAuth: The safe “Sign in with Google/Microsoft” flow apps use.
Dry-run: Practice mode—shows what would happen without sending for real.
Timeline (Hub tab): Live feed of recent agent runs and bridge actions—not the same as Chronologies.
Chronologies (Hub tab): Merged, dated event log you review, link, and export (built by Timeline builder).
Timeline builder: The agent that merges mail/docs/feeds into a chronology instance.
Quick note: Preset review sentences in Chronologies—pick one to fill the review box, then save.
Research Library: Model Pile sidebar that saves compare, critique, debate, and vision sessions for reuse.
Project Research Library: Research Library linked to a SharePoint project folder (sources, exports, consensus).
SharePoint organizer: Hub agent that connects to OneDrive/SharePoint via Microsoft Graph—can organize files and host project research folders.
I want to…: Intent picker on the Agents tab—shortcuts to Hub, Research Library, AgentTasker, and Marketing paths.
Slug: Short folder name (lowercase, hyphens), e.g. acme-mailbox.