vane
StarContact us中文

Multimodal context.Ready for your agents.

Connect enterprise content and business data. Retrieve, analyze and retain the knowledge each task needs, with less repeated processing and lower context costs.

Request a demo →View integration options
A drink at the campsiteUnderstanding, with evidence
00:0000:21

Outdoor campsite

A drink comes together at the campsite: pour the ingredients, then start blending.

Interactive example · curated understanding

See Vane Context in action.

Follow a marketing task from enterprise data to cited context, then see how it supports other agents.

From enterprise data to agent-ready context.

Connect your sources, keep knowledge current and deliver cited context through tools and Skills.

Vane Context architecture. Enterprise data feeds preparation and a shared data space. Context services retrieve and deepen understanding, publish knowledge back, and serve business agents through MCP Tools and Skills. Vane Data supports processing; human oversight spans the product.

Enterprise data

Files & media · Business tables · External semantic definitions

Vane Context

Processing and updates

Sync changes → extract and analyze → embed and index

Data space

Files & media · Tables · Knowledge · Semantic definitions · Search indexes

Business tables and external definitions can stay in their source systems and be bound here.
Retrieve existing knowledgeSave analysis results

Context service

Search & business queries · On-demand understanding and media processing

Context AgentDelegate to a Context Agent
↑ Request ↓ Results with citations

Agent integration

MCP Tools · Skills

MCP tools expose capabilities; Skills guide the agent.

Vane Data

Supports preparation and online processing: native media, UDFs, AI Functions, relational queries and vector search.

Management and observability

Verify sources · Manage semantic definitions · Troubleshoot · Monitor usage and costs

↑ Request ↓ Result

Business agents

Use the context to create, decide and act.

Update searchable knowledge as your data changes.

Analyze relevant content when a task needs more detail.

Save the results as knowledge for future tasks.

Connect multimodal content, business data and semantics.

Link files, media and tables to business models, relationships and metric definitions so agents can use them in context.

Original campsite video.

comp_001

Original video, transcripts, action clips and source references.

creative_id: comp_001View the source clip →

ad_daily

creative_id
comp_001
Business fields
date / impressions / clicks / conversions

Campaign performance

ad_daily.creative_id = creatives.id

Conversion rate = conversions / clicks. Compare the same reporting period and scope.

Link the source clip to business records through its asset ID.

Example data connections. Campaign metrics are supplied by the relevant business records.

Connect local files, object storage, business databases, data catalogs and external semantic layers.Discuss your data sources →

Automate data preparation. Analyze on demand.

Keep knowledge up to date as data changes. Retrieve what is already known, analyze the missing details and save reusable results.

Continuous processing

Extract content with source locations, and maintain embeddings, indexes and source versions.

Source referencesInitial analysisEmbeddings & indexes
Agent request

Find footage that shows portable use, with evidence.

  1. Retrieve the relevant part

    Locate candidate clips using existing indexes.

  2. Analyze the missing details

    Analyze only the relevant clip.

  3. Return task-relevant context

    Return the finding and its original reference.

Original product-action window.
New analysis

Blending at the campsite

The footage shows ingredients being blended on the outdoor table.

comp_001 / 00:05–00:07

Validated analysis results are saved to the data space for future retrieval and reuse.

Knowledge accumulates. Each task retrieves what it needs.

One connection. Tools and Skills, on demand.

Connect your existing agent through MCP. Discover authorized capabilities and task-specific guidance from the service as you work.

  1. 01

    Connect your MCP client

    Choose the data spaces and capabilities to share. Copy the connection settings into your agent’s MCP client.

  2. 02

    Verify tools and guidance

    Ask the agent to check the connection and discover the available tools and Skills. Only authorized, published guidance is returned.

  3. 03

    Start with a business task

    Give the agent your task, known facts and constraints. It reads relevant Skills when needed and retrieves or prepares context through the tools.

View an example MCP configuration
mcp-config.json
{
  "mcpServers": {
    "vanecontext": {
      "url": "<MCP_ENDPOINT_URL>",
      "headers": {
        "Authorization": "Bearer <CONNECTION_TOKEN>"
      }
    }
  }
}

Use the endpoint and token from your service. The exact configuration format depends on your MCP client.

See it with your agent →
Guidance that fits the task

The right Skill, when your agent needs it.

The service exposes guidance and scenario Skills through the same connection. Your agent discovers their purpose and version, then reads only the instructions relevant to the task.

Discoverlist_context_skillsPurpose · scope · version
Read as neededread_context_skillSKILL.md + references

Use tools directly

Let your agent search, query, inspect evidence and request authorized processing within its own workflow.

Delegate context research

Let a Context Agent investigate missing enterprise information and return findings, sources and gaps. Your agent continues the business task.

Built-in and external agents use the same guidance. Clients with MCP Skills support can discover Skills natively; other clients use the guidance tools. Available actions always follow the service’s permissions.

Visibility into sources, processing and cost.

Verify results, manage semantic definitions, track tasks and monitor model usage and costs in one workspace.

Verify results against the source

Open the original image, document page or media time window behind a result. Compare the source with the interpretation, and distinguish observed facts from inferred conclusions.

Manage processing and semantic definitions

Choose processing channels and models. Review business models, relationships and metric definitions before publication, and manage the knowledge your agents can use.

Track tasks and troubleshoot issues

Follow processing and context tasks. Locate the affected file, failed step and reported cause, then retry the work that needs attention.

Monitor usage and costs

Review token usage, estimated costs, reuse and compression by model, task and processing stage. Missing usage or pricing is clearly labeled.

Agent solutions built around your business tasks.

Combine data connections, task-specific processing, a semantic layer and agent skills across six business use cases.

VANE CONTEXTExample scenario
Business context for your agent

Outdoor creative references

  • Relevant clip: a drink made at the campsite.
  • Visible action: pour, add ice, blend.
  • Support product claims with approved specifications and brand information.
Original media · example brief and campaign data

Connect product facts, brand guidelines, creative assets and campaign data to give marketing agents verifiable references and evidence for planning.

Explore this use case
Explore all solutions