Introduction to TextCortex Flows – TextCortex Help Center

Bring TextCortex agentic capabilities into workflow automation.

Flows is TextCortex’s new workflow automation product, built to extend the TextCortex Platform from interactive AI conversations into background execution. It enables teams to build sophisticated multi-step automations in a visual, node-based editor, using the exact same agentic capabilities and resources they already know from TextCortex, now combined with deterministic workflow components for structure, control, and repeatability.

Flows is not a replacement for the TextCortex Platform. It is a complementary product. The platform remains the place for real-time collaboration with AI, where users interact with agents, ask questions, and generate outcomes in the moment. Flows takes that same intelligence and operationalizes it. Once a workflow is triggered, it can run automatically in the background, researching, reasoning, transforming data, making decisions, and taking action without requiring a human to guide each step.

What makes Flows especially powerful is its ability to combine autonomous AI workflows with structured deterministic processes.

Teams can use their existing agents where work requires judgment, synthesis, and adaptability, such as analyzing research, identifying patterns, or deciding what matters in a large set of information.

At the same time, they can rely on deterministic components like conditional logic, loops, data transformations, API calls, and security guardrails to ensure workflows remain controlled, reliable, and production-ready. In other words, Flows gives organizations the best of both worlds: intelligence where flexibility matters, precision where consistency matters.

A clear example is weekly competitive intelligence monitoring. With Flows, a team can build a workflow that runs every Monday morning and automatically researches the latest industry trends across news sites, research papers, and competitor announcements. It can analyze key developments, identify what is relevant to the business, compile everything into a structured digest, and email it directly to the product team or add the report to a knowledge base. The entire process happens in the background, turning a time-consuming manual task into a repeatable system that delivers actionable insights before the workday even begins.

Scheduling Flows

Flows can be set to run automatically on a recurring schedule. Once configured, a workflow triggers at the time and frequency you define, runs completely in the background, and delivers its output without anyone needing to start it manually.

You can schedule workflows to run hourly, daily, weekly, or on a custom cadence.

Use cases

Weekly competitive monitoring. Every Monday morning, a workflow searches for new announcements from key competitors, passes results to an agent for analysis, and emails a structured digest to your product team before the workday starts.
Daily news briefing. A workflow runs each morning, searches for industry news across your chosen topics, filters for relevance, and drops a summary into a knowledge base your agents can reference throughout the day.
Regular knowledge base updates. A workflow crawls a set of URLs every week (documentation sites, research portals, competitor blogs) and indexes new content into a knowledge base automatically, keeping it current without manual uploads.
Recurring reporting. An end-of-week workflow pulls data from an external source via HTTP Request, processes it through an agent, and sends a formatted performance summary to a distribution list every Friday afternoon.

Scheduling works especially well when paired with the Send Email component for delivery, or the Add to Knowledge Base component for accumulating intelligence over time.

Because Flows is built visually, teams can orchestrate complex automations step by step. A workflow might ingest URLs or other inputs, parse and structure content, pass it to an agent for analysis, branch using If/Else logic, store results in a knowledge base, and trigger downstream actions such as notifications or emails. This makes Flows suitable for a wide range of use cases, from market research and reporting to knowledge operations, internal workflows, and process automation across teams and systems.

It allows organizations to take the same agentic capabilities that power the TextCortex Platform and embed them into automated workflows that are intelligent, repeatable, and scalable.

Main Components

Webhook (HTTP Request)

What it does

The HTTP Request component sends a request to any external API and pulls the response back into your workflow. Use it to connect Flows to tools, databases, or services that TextCortex doesn't have a built-in integration for.

Fields

URL: The API endpoint you're sending the request to.
Method: The type of request. GET retrieves data. POST sends data. PUT and PATCH update existing records. DELETE removes them.
Headers: Key-value pairs the API needs to process your request, like authentication tokens or content type declarations.
Body: The data payload you're sending, typically used with POST or PUT requests.
Authentication: How you prove you're allowed to access the API. Options vary by provider (API keys, Bearer tokens, Basic auth).
Timeout: How long Flows waits for a response before stopping. Default is 30 seconds.

Outputs

Response: The data the API sends back.
Status Code: A number indicating whether the request succeeded (200 means OK, 4xx means client error, 5xx means server error).

Use cases

Pulling fresh CRM data from a tool TextCortex doesn't natively connect to, then feeding it into an agent for analysis.
Sending a structured summary generated by an agent to an internal dashboard via POST.
Triggering an action in an external system (like creating a ticket or updating a record) at the end of a workflow.

URL (Web Crawler)

What it does

The URL component fetches content from one or more web pages and can follow links from those pages automatically. Think of it as a crawler you can point at a starting URL and tell how far to go.

Fields

URLs: The starting pages for the crawl.
Depth: How many "clicks" away from the starting page the crawler will follow links. Depth 0 means just the page you specified. Depth 1 means the starting page plus every page it links to. Depth 2 goes one level further. Keep this number low unless you have a specific reason to go deep.
Other options (output format, timeout, etc.) behave the same as the Browserless URL component.

Output

Fetched Pages: All the content collected across the crawl, returned as structured text.

Use cases

Crawling a competitor's blog index page (depth 1) to pull all recent posts into a content analysis workflow.
Ingesting all pages of a public knowledge base or documentation site into a TextCortex knowledge base.
Building a monitoring workflow that checks a multi-page resource regularly and flags new content.

Web Scraper (Browserless URL)

What it does

The Browserless URL component fetches the full content of one or more web pages using a headless browser. Unlike a standard URL fetch, it can handle pages that load content dynamically with JavaScript, and it's built to work around common bot-detection measures.

Important: This component is for accessing publicly available content. It's not designed to bypass paywalls or access content that requires a subscription or login.

Fields

URLs: The web addresses you want to fetch. You can add multiple.
Output Format: How the page content gets returned (plain text, HTML, markdown, etc.).
Content Extraction: Whether to pull the full page or just the main body content, stripping navigation and footers.
Use Residential Proxy: Routes the request through a residential IP address, which helps avoid blocks on sites with stricter access controls.
Timeout: How long to wait for the page to load before giving up.
Max Retries: How many times the component tries again if the first attempt fails.
Continue on Error: If enabled, the workflow keeps running even if one URL fails to load.

Output

Fetched Pages: The extracted content from each URL, ready to pass into an agent or transformation step.

Use cases

Scraping competitor product pages weekly to track positioning changes, then sending a digest to Slack.
Pulling the full text of a news article for an agent to summarize, even when the page renders content dynamically.
Ingesting public documentation or whitepapers into a knowledge base automatically.

Web Search

What it does

The Web Search component runs a Google search and returns results directly inside your workflow. It uses SerpAPI under the hood and supports a range of search types beyond standard web results.

Fields

Search Query: What you're searching for. This can be static text or a dynamic value passed from an earlier step.
Search Mode: The type of search to run. Options include Web, News, Scholar, Wikipedia, YouTube, Reddit, X (Twitter), and Patents.
Site Filter: Restrict results to a specific domain (e.g., techcrunch.com).
Max Results: How many results to return.
Fetch Page Content: If enabled, the component goes one step further and retrieves the actual content of each result page, not just the title and link.
Language / Country: Narrow results by region or language.
Timeout: How long to wait before stopping the search.

Output

Search Results: A structured list of results, including titles, URLs, and snippets (and full page content if that option is on).

Use cases

Running a weekly News search for your company name or key competitors, then passing results to an agent to flag anything relevant.
Using Scholar mode to pull recent research on a topic and feed it into a report-writing workflow.
Searching Reddit or X for mentions of a product category to surface early signals for a market research workflow.

Add to Knowledge Base

What it does

The Add to Knowledge Base component takes content from your workflow and saves it directly into one of your TextCortex knowledge bases. It's how you turn workflow outputs into persistent, searchable knowledge that your agents can reference later.

Fields

Input Type: What you're uploading. Options are Text (raw content), File (a document), or URL (a web page to fetch and store).
Content: The actual text, file, or URL to ingest, typically passed in from an earlier step.
Knowledge Base: A dropdown showing all the knowledge bases in your TextCortex workspace. Just select the one you want to add content to.

Use cases

At the end of a research workflow, automatically saving the agent's output report into a knowledge base so future agents can reference it.
Storing scraped web content (from the URL or Browserless URL component) into a knowledge base for ongoing competitive monitoring.
Building a workflow that processes incoming documents and indexes them automatically, so your team's knowledge base stays current without manual uploads.

Retrieve from Knowledge Base

What it does

The Retrieve from Knowledge Base component searches one of your TextCortex knowledge bases and pulls back the most relevant content for a given query. Use it when you want a workflow to reference your internal documents, past reports, or indexed content before an agent produces its output.

Fields

KB: A dropdown showing all the knowledge bases in your TextCortex workspace. Select the one you want to search.
Query: The question or topic you're searching for. This can be static text or a dynamic value passed in from an earlier step.
Number of Results: How many matching chunks to retrieve. Default is 5. Lower it for tighter, more focused context; raise it if the agent needs broader coverage.

Outputs

Retrieved Text: The matched content as plain text, ready to pass directly into a Prompt or Agent component.
Results: The full structured result set, including metadata like source names and relevance scores.

Use cases

Retrieving relevant internal guidelines before an agent drafts a client-facing response, so the output stays on-brand and accurate.
Pulling matching past research reports into a workflow before generating a new analysis, so the agent builds on what already exists rather than starting cold.
Searching a product knowledge base for relevant specs or FAQs, then combining the results with a web search before sending a competitive briefing.

Agent

What it does

The Agent component is where the AI thinking happens. It runs a multi-step query using the model, tools, and knowledge you configure, then returns a result you can pass to the next step.

You can connect an existing TextCortex agent (with all its settings already baked in) or build a fresh one directly inside the workflow with custom instructions, tools, and knowledge bases.

Fields

Agent Selection: Choose an existing agent from your TextCortex workspace, or start from scratch with a blank configuration.
Model: Which AI model the agent runs on. Different models trade off speed, reasoning depth, and cost.
Tools: What the agent can do during its run. Options include web search, data analysis, code execution, knowledge base retrieval, and more. You can stack multiple tools together.
Knowledge Bases: Attach one or more knowledge bases so the agent can pull from your internal content when forming its response.
Text Input: The instruction or query the agent receives, usually passed in dynamically from an earlier step in the workflow.

Use cases

Receiving a batch of scraped competitor pages and analyzing them for positioning changes, pricing updates, or new features.
Taking structured data from an HTTP Request and producing a written summary or recommendation based on your internal guidelines.
Running a research workflow where the agent searches the web, cross-references your knowledge base, and writes a final briefing.

Prompt

What it does

The Prompt component is a plain text field for passing instructions into the next step of your workflow. It doesn't run an agent itself. It shapes what the agent or downstream component receives.

Use it to write reusable instruction templates that combine static text with dynamic values from earlier steps.

Fields

Prompt Input: Free-form text. You can type fixed instructions and insert dynamic variables from previous steps using the variable picker.

Use cases

Writing a reusable instruction like "Summarize the following content in 3 bullet points for a non-technical audience: [content from previous step]."
Constructing a prompt that combines a user-defined query with context pulled from a knowledge base before sending it to an agent.
Chaining multiple prompts to progressively transform content across several agent steps.

Text Input

What it does

The Text Input component is the entry point for manual or dynamic text entering a workflow. It defines what gets fed in when the workflow runs.

Fields

Text: The value you want to pass into the workflow. Can be a fixed string or, in triggered workflows, a dynamic value supplied at runtime.

Use cases

Setting a search term or topic that gets passed to a Web Search component downstream.
Defining a document title, company name, or any other variable that shapes what the rest of the workflow does.
Testing a workflow manually by typing in sample input before connecting a live trigger.

Text Output

What it does

The Text Output component displays the final text result of a workflow run. It's typically the last node in a chain, showing you what the workflow produced.

Fields

Output: The text value passed in from the previous step. Read-only at runtime.

Use cases

Previewing the output of an agent summary before deciding to route it elsewhere (email, knowledge base, etc.).
Showing the result of a data transformation or parsing step during workflow testing.
Acting as the final display node in workflows that are run manually and reviewed on screen.

Mapping (Parser)

What it does

The Parser component extracts specific pieces of text from structured content using a template you define. You map a column or field from the previous step, then tell the parser what to pull out of it so the result can be cleanly passed to the next step.

Fields

Input: The content field or column from the previous step you want to parse.
Template: The pattern or structure the parser uses to identify and extract the relevant portion of the text.

Use cases

Extracting just the article body from a scraped page before sending it to an agent, trimming out metadata and navigation text.
Pulling specific values (like a price, a date, or a product name) from a structured text response.
Cleaning up agent output before it gets written to a knowledge base or sent via email.

HTML Parser

What it does

The HTML Parser component extracts specific data from raw HTML using CSS selectors. Where the standard Parser works on text patterns, the HTML Parser works on page structure, so you can target exactly the element you want.

Fields

HTML Input: The raw HTML content, usually fetched by the Browserless URL or URL component.
CSS Selector: The selector that identifies the element(s) to extract. For example, h1 pulls all top-level headings, .price pulls elements with the "price" class.

Use cases

Extracting just the headline and body text from a news article page, ignoring ads and sidebar content.
Pulling a list of product names and prices from a competitor's pricing page.
Targeting a specific section of a documentation page for indexing into a knowledge base.

Data Operations

What it does

The Data Operations component lets you manipulate a data object between workflow steps. Instead of sending a raw blob of data to an agent or output, you can reshape it first so only the right information moves forward.

Available operations

Select Keys: Keep only the fields you want from a data object, dropping everything else.
Count: Return the number of items in a list or object.
Literal Eval: Convert a text string that looks like structured data (like a list or dictionary) into an actual data object the workflow can work with.
Combine: Merge 2 or more data objects into one.
Filter Values: Keep only items that match a condition you set.
Append or Update: Add new fields to an existing object, or update the value of an existing field.
Remove Keys: Delete specific fields from a data object.
Rename Keys: Change the name of a field without changing its value.

Use cases

After an HTTP Request returns a large API response, using Select Keys to isolate just the fields your agent needs.
Combining results from 2 separate Web Search steps before passing everything to an agent for synthesis.
Filtering a list of search results to only items published after a certain date before analysis.

If-Else

What it does

The If-Else component routes your workflow down different paths based on a condition. If the condition is true, the workflow goes one way. If it's false, it goes another. This is how you build decision logic without needing to write code.

Fields

Input: The value being evaluated, usually the output of a previous step.
Operator: How to compare the input. Options include equals, contains, starts with, ends with, is greater than, and more.
Comparison Value: What you're comparing the input against.
Iterations: For list-based inputs, the If-Else can evaluate each item individually and route each one based on its own value.

Use cases

Checking whether a web search returned results before passing them to an agent. If results are empty, route to a fallback step instead of failing silently.
Routing an agent's output to different email recipients based on a keyword in the content.
Processing a list of scraped pages and separating ones that mention a specific competitor from ones that don't.

Send Email

What it does

The Send Email component delivers workflow output directly to one or more email inboxes. It supports HTML formatting, so the emails it sends can be structured and readable, not just raw text dumps.

Fields

To: One or more recipient email addresses.
Subject: The email subject line. Can include dynamic values from earlier steps.
Body: The email content. Supports HTML, so you can format with headings, bullet points, and links.

Use cases

Sending a weekly competitive intelligence digest to your product team every Monday morning, automatically compiled and formatted by an agent.
Pinging a colleague when a workflow detects a specific trigger, like a competitor announcing a new product.
Delivering a structured report to a distribution list at the end of a research or monitoring workflow.