Task Agent × Visual Workflow

Use an Agent for complex work. Use a workflow for repeatable processes.

Describe the outcome and add your materials. A Task Agent plans the work, uses tools, and organizes reports, datasets, webpages, and other deliverables in your workspace; some third-party media results are delivered by link. Use a visual workflow for processes you need to run repeatedly.

New accounts receive 5,000 credits. No card required.

Public task · European accelerator and incubator database CASE WALKTHROUGH
Collect accelerators and incubators open to European teams and organize the results in a CSV
Cycle 8 / 8 2 files delivered 47 organization records

Split the goal into research and CSV preparation, collecting websites, social profiles, locations, and verifiable funding details.

Workspace

Public cases

Real tasks and deliverables

These tasks were shared publicly by their creators. Open a case to inspect the published steps, tool records, and deliverable files.

Platform at scale

200+model configurations
across text · image · video · audio
100+visual workflow nodes
across 17 functional categories
29built-in Agent tools
for files · search · terminals · sub-tasks
2ways to work
Task Agent · visual workflow

Two ways to work

Open-ended work and repeatable processes

Use a Task Agent when the path may change and the work needs judgment. Build a visual workflow when the steps are known and need to run repeatedly. Both can connect models, files, data, and external tools.

Task Agent

Give a goal and materials; the Agent plans, executes, and corrects itself. Built for complex work without a fixed path.

  • Review messages, actions, observations, and tool results by cycle
  • Use files, knowledge bases, workflows, MCP, and sub-agents
  • Add context during a task and receive completion notices

Visual Workflow

Define inputs, processing, and outputs with nodes for processes with known, repeatable steps.

  • Drag to connect models, data, media and output nodes
  • Official & community templates, or start blank
  • Run on a schedule, keep records, and share when needed

Agents can invoke configured workflows and MCP tools. Workflows can also be added to an MCP Server for compatible clients to call.

Collaboration across your organization

Use organization-shared files directly in Agents and workflows

Admins can manage members and nested department structures. Share files and folders with selected members, departments, or the whole organization, with view-only or collaboration access. Authorized shared files can be used in Agents and workflows, and Agent-generated files can be saved back to shared folders where you have collaboration access.

  1. Organizations, departments, and members

    Manage members, invitations, and multi-level departments so sharing follows the way your team is structured.

  2. Share materials by scope

    Share files and folders with selected members, departments, or the whole organization, with view-only or collaboration access.

  3. Use shared files in Agents and workflows

    Use shared files as Agent reference material or workflow inputs, then save Agent-generated files back to shared folders where you have collaboration access.

Choosing a tool

VectorVein, Codex, or Claude Code?

Choose based on both the final deliverable and what your team needs to share over time. VectorVein lets teams use organization-shared files in Agents and workflows, then save outputs back to shared folders; Codex and Claude Code center collaboration on repositories and software-development processes.

Short answerChoose VectorVein when the main deliverable is a report, dataset, webpage, or media result, or when your organization needs to reuse shared files and standardize repeatable processes as workflows. Choose Codex or Claude Code when the main deliverable is code; their scopes overlap, so base the choice on your existing account, model preference, IDE/CLI workflow, and team conventions.
Primary positioning and organization collaboration models of VectorVein, Codex, and Claude Code
DimensionVectorVeinCodexClaude Code
Primary workResearch, content, data processing, and other multi-step tasks with a defined deliverable.Build, debug, refactor, and review code in a software repository, and run tests.Build, debug, refactor, and review code in a codebase, and run tests.
Primary contextWebpages, Office and PDF files, datasets, knowledge bases, media, and external tools.Code repositories, terminals, IDEs, Git, and development tools.Codebases, terminals, IDEs, Git, and development tools.
Organization and team collaborationBuilt-in organization, department, member, and invitation management. Share files and folders with members, departments, or the whole organization, set view-only or collaboration access, and use shared files directly in Agents and workflows.Team collaboration mainly centers on repositories, shared development guidance, code review, pull requests, and related engineering processes. ChatGPT Business and Enterprise workspaces also provide admin controls.Team collaboration mainly centers on repositories, project settings, GitHub workflows, pull requests, and related engineering processes. It supports team-shared settings and organization-managed configuration.
When to choose itYour main deliverable is a report, dataset, webpage, or media result, or your organization needs to reuse shared files and standardize repeatable processes as workflows.Your main deliverable is code and you prefer OpenAI models, or want to work through the desktop app, IDE, CLI, cloud tasks in the web app, or SDK.Your main deliverable is code and you prefer Claude models, or want to work through the terminal, IDE, desktop app, browser, GitHub Actions, or Agent SDK.
Typical outputsReports, datasets, webpages, media results, workspace files, and reusable workflows.Code changes, diffs, test or command results, code-review feedback, commits, and pull requests.Code changes, diffs, test or command results, code-review feedback, commits, and pull requests.

Use them togetherThey can divide the work: use Codex or Claude Code to build and maintain code, and VectorVein to run and reuse business processes. Claude Code can call published workflows through VectorVein's current SSE MCP endpoint. Codex remote MCP currently requires Streamable HTTP, so use a REST API or an adapter for now.

Features and access options can change. See the official documentation: CodexCodex MCPClaude CodeClaude Code SettingsClaude Code MCP

Core capabilities

Run code, work with files, and deliver usable outputs

Review task steps and tool calls, then find reports, webpages, datasets, and other deliverable files in the workspace. Some third-party media results are delivered by link.

Process files and run analysis

Agents can run Bash and Python and read or write files in an isolated cloud environment that shuts down when idle.

Move complex work forward in parallel

A main Agent can split work into sub-tasks so research, data processing, and report writing can proceed in parallel.

Retrieve material with source file paths

Connect Office documents, PDFs, and cloud files, then retrieve relevant content with its original file path for review and citation.

Run by time or task status

Schedule runs or start follow-up tasks and workflows when a task changes state.

Reuse steps that have been verified

Agents can invoke configured workflows and external tools. Repeatable steps can be organized as visual workflows.

Bring your own API key

Use your own API keys without spending platform credits on model calls

Calls made with your own OpenAI, Azure, Anthropic, Vertex, or Bedrock keys do not use platform credits; your model provider bills those calls under its own terms. You can choose whether to enable platform-key fallback, and fallback calls use platform credits.

  • Choose your own key or a platform key for each model
  • Configure your own keys and platform-key fallback separately
  • See an itemized breakdown of credit usage for model calls and workflow runs
OpenAIgpt-5 · o4 · gpt-4o
Own-key calls: 0 platform credits
Anthropicclaude-opus · claude-sonnet
Own-key calls: 0 platform credits
Vertex / Azure / BedrockEnterprise cloud model channels
Own-key calls: 0 platform credits

Ways to access VectorVein

Start work from the devices and systems you use

Start tasks from the web workspace, browser extension, supported mobile apps, or REST APIs. Workflows can also be published as web apps or called by compatible MCP clients.

Browser extension

Send selected text, links, or page content to enabled workflows from Chrome, Firefox, or Edge.

iOS App

Available in selected App Store regions for viewing and running tasks and workflows on iPhone.

Open in the App Store

Developer API

Use an API key to manage and run workflows and Task Agents through REST APIs.

Manage API keys

Connect and publish with MCP

Connect MCP services to Task Agents, or expose selected workflows through a configured MCP Server for compatible clients.

Publish as a VApp web application

Publish a workflow as a web app and control access with keys, usage limits, credit caps, and expiration dates.

Create a VApp

Before you start

Frequently Asked Questions

Looking for setup and usage details? Read the docs

How does pricing work?

Creating and configuring Task Agents and editing workflows do not use credits. Runs consume credits based on the platform models and paid tools used. Calls made with your own third-party model keys do not use platform credits; your model provider bills those calls under its own terms. View model credit pricing

Who is VectorVein for?

VectorVein fits research, content, operations, and data work that involves finding sources, processing files, using multiple tools, and keeping reviewable steps and deliverables. Fixed processes can be built as visual workflows. For a one-off short answer, a standard chat product may be simpler.

How is VectorVein different from Codex and Claude Code?

Consider both the deliverable and what your team needs to reuse over time: choose VectorVein for reports, datasets, webpages, or media, or when your organization needs to reuse shared files and standardize repeatable processes as workflows; choose Codex or Claude Code for code changes, tests, and review around a codebase. They can also be used together. See the full comparison

Are my tasks and workspace files public by default?

No. New tasks are not shared and do not appear in the community by default. When you explicitly enable sharing in task details, anyone with the link can view the task description, execution records, tool calls, and workspace files. Enabling Public also lists the task in the community. Sharing a parent task also makes its subtasks accessible by link, so review and remove sensitive content before sharing.

Does using my own model API key consume platform credits?

Successful third-party model calls made with your own key do not use platform credits; your model provider bills you directly. You can choose to use only your own key or fall back to a platform key when it is unavailable. Only fallback calls use platform credits.

Can I try for free?

Yes. New accounts receive 5,000 credits and can try Task Agents and workflows without adding a payment method first.

Is there an open-source edition?

Yes. Download the open-source desktop edition from GitHub Releases and run it locally. It focuses on visual workflows; Task Agents are currently available on the hosted VectorVein platform.

Start with a task you need done.

Describe the outcome and add your materials. New accounts receive 5,000 credits with no payment method required.

Create a task for free