Browse documentation

Quick start

Go from an empty workspace to your first automated result in about 10 minutes.

This is the guided happy path for a brand-new workspace. In a handful of steps you'll bring in some data, visualize it, track work against it, automate a step, put an agent to work, and finish by wrapping everything up in a Space. Each step links to the deeper guide if you want more detail.

You don't have to do these in order forever, but the first time through, following them top to bottom gives you a complete tour of how the pieces of SPARK connect.

Before you start

You'll need access to a SPARK workspace. If someone on your team already set one up, ask them to invite you; if you're the first person in, you'll land in an empty workspace ready to fill. Have a small CSV handy: a list of counterparties, positions, funds, or anything you already track in a spreadsheet works perfectly for this walkthrough.

1. Get oriented

When you open SPARK you land on the workspace's New Thread page: a greeting, the composer to hand work to an agent, and your recent items.

Good afternoon, ClaireAsk Spark or name an agent…RecentsOvernight NAV moves above 2%5 minutes agoNAV movements2 hours agoCheck the Kestrel Global Equity NAVyesterday
The New Thread page, where a workspace opens. The composer hands work to an agent.

The left sidebar is your map. At its top, three icon tabs switch what it shows: Threads (your conversations with people and agents), Ecosystem (what the workspace holds) and Settings (configuration). You'll spend this quick start mostly on Ecosystem, whose panel lists Studio, Tasks and Canvas, then Workflows, Data and Files, which Less folds under More. The button at the very top of the sidebar shows your workspace name; click it to switch workspaces if you belong to more than one.

For a full tour, see Navigating the workspace.

2. Bring in your data

Data is the foundation everything else builds on, so start there.

  1. On the Ecosystem tab, click Data (under More if it is folded).
  2. Click New at the top right of the Data page.
  3. In the New data object dialog, give it a clear Name (for example, Counterparties), and pick a Group if you like.
  4. Under Attributes, in Upload a CSV, click Choose a file and pick your CSV, then click Extract attributes with AI: SPARK reads the columns and proposes a name and a type for each, for you to review. Leave Import the file's rows after creating ticked so the rows come in too.
  5. Click Create.
CancelCreateNew data objectName the data object. You can also create its attributes from a CSV file.NameCounterpartiesGroupMainAssign this data object to a group (optional)AttributesCancel uploadOptional. Create them from the columns of a CSV file.counterparties.csvExtract its columns as attributes, then review them before creating.Extract attributes with AIImport the file's rows after creating
The New data object dialog: name it, then build its attributes from a CSV.

SPARK creates the object and loads your records. To start from one of SPARK's presets instead, open the arrow beside New and choose Create from template; see Data object templates. If you'd rather import into an object that already exists (with AI-assisted column mapping), see Add, edit & import records. For the object model, see Create a data object.

3. Build your first canvas

A canvas is a board of widgets: tables, charts and KPI tiles built directly on your data.

  1. On the Ecosystem tab, click Canvas, then New. There is no dialog: SPARK creates a canvas called Canvas and opens it in edit mode, with the Widgets library open beside it. Click its title to rename it.
  2. Click New widget in the library, pick what kind of widget to make, such as a Default table pointed at your data object, then a Chart or a KPI to surface a number that matters. Each opens its editor, where you pick its data and how it shows. See Create a widget.
  3. Click Save and close.
24Counterparties1,250.00Total limit (m)Counterparties by country0487GB5FR4LU3DE3NL2CHCounterpartiesEvery counterparty, with its rating and its limit.NameCountryRatingLimit (m)Contoso BankGBA250.00Fabrikam SecuritiesFRBBB100.00Contoso CustodyLUA180.001-10 of 24Page 1 of 3
A canvas with KPIs, a chart and a table, all reading from one data object.

To edit the canvas again later, choose Edit canvas & context from its ⋯ menu, or press ⌘E (Ctrl+E on Windows). See Build a canvas to go further.

4. Create a task and upload a file

Tasks track work and connect it to the data it relates to; files hold the documents that work depends on.

  1. On the Ecosystem tab, click Tasks, then New.
  2. In the New task dialog, type a Title (for example, Review Q3 counterparty exposure) and an optional Description, pick yourself with the assignee selector in the row of properties, then click Create.
  3. Open the task and, in its Canvas section, click + (Add canvas) and pick your canvas, so the board travels with the work.
  4. Files live on their own page. Click Files (under More if it is folded), click New, add a document in the Upload Files dialog, and click Submit. You can also drop a file straight onto the Files page.

See Create and assign and Upload and organize.

5. Automate a step with a workflow

A workflow turns a repeatable operation into something that runs on its own, from a trigger (what starts it) through one or more actions (what it does).

  1. On the Ecosystem tab, click Workflows (under More if it is folded), then New. There is no dialog: SPARK creates a workflow called Workflow and opens its editor.
  2. Build it from a trigger and one or more actions. For a sign-off step, add a Human Review action; pending items then wait under Interventions, where a reviewer can approve or reject them.
  3. A new workflow starts as a draft, so click Save, then Activate in the editor's toolbar when it's ready.

See Build a workflow and Human-in-the-loop.

6. Put an agent to work

Every workspace already has Spark, a built-in agent with access to SPARK's tools, so you can start straight away: type a request in the composer on the New Thread page and send it. A message that names no agent goes to Spark, and the thread opens in your direct message with it. See The Spark agent and Start a thread.

To build an agent of your own, with its own instructions, model, tools and skills:

  1. On the Ecosystem tab, click Studio (or open the Settings tab, where Agents is the first page under Agent Studio), and click New. A blank agent opens in the editor, with no dialog. The arrow beside New also offers From template.
  2. Give the agent a name, a model and instructions and, optionally, the tools and skills it should have, then click Create agent. See Create and configure.
  3. Open New Thread and name your agent in the composer, or pick it under Direct Messages on the Threads tab. An agent answers only a message that names it, except in your own direct message with it.

Because agents run inside your workspace, they can already see your data object, canvas, and tasks, with no copying and pasting. Try prompts like:

  • "Summarize the records in my Counterparties object."
  • "Which of my open tasks are unassigned?"
  • "Draft a note explaining the KPI on my canvas."

7. Wrap it all up in a Space

You've now built a handful of separate pieces. A Space bundles them into one use case so your team has everything for this work in a single place.

  1. On the Ecosystem tab, hover the Spaces section of the sidebar and click its + (New space). There is no dialog: SPARK creates the Space and opens its page. Click its name to rename it and say what it is for.
  2. On the Space's page, click + (Add item) in the page header, then Feature, to add the data object, canvas, tasks and agent you built.
Counterparty monitoringCounterparty exposure and limits, checked every month.ShareTasksReview Q3 counterparty exposureLimits against the latest exposureNov 14DataCounterpartiesEvery counterparty, with its rating and its limitCanvasCounterparty overviewCounterparties by country, their limits, and the…ThreadsWhich counterparties are near their limit?2 hours ago
A Space's page: what it holds, by feature, and the threads about it.

From now on, opening the Space gives everyone the same focused starting point; it shows on the Ecosystem page and at the foot of the Threads panel, and the threads started about it gather on its page. See Create a Space to go deeper.

What you've built

In one pass you've touched every core object in SPARK and seen how they reinforce each other: data feeds the canvas, tasks and files track the work, a workflow automates a step, an agent acts on the whole picture, and a Space ties it all together for one use case.

Where to go next