BaseCiteSlow Data AuroraRequest access
Customer-submitted / unverified / bounded context

Bounded context for AI and agents

MCP and API clients retrieve scoped context labels while private originals stay outside public retrieval.

AI access flow: request scoped context, preserve labels, avoid raw files, prevent dataset dumps.

Request accessExplore platform
Customer-submittedUnverifiedNot evaluatedBounded contextNo raw downloadNo bulk export
Customer-submitted
Private originals
Bounded context
Controlled retrieval
Bounded context is activeOnly the right information. In the right way.
System healthy
Customer-submittedIntake
Private originalsProtected
Bounded contextPrepared
Controlled retrievalDelivered
Private originalsProtection fieldBounded context
Platform

Enter the protection field.

Uploads are treated as customer-submitted and unverified. The product flow records consent, handling status, correction paths, and bounded context without turning submitted materials into claims.

UploadCustomer and Solution Company entry points keep consent and scope visible.
Private originalsOriginal materials stay separated from AI-readable context surfaces.
CorrectionsCorrections append new records instead of silently overwriting history.
Bounded contextOnly limited context capsules cross the protection field.
Solutions

Context stream, not file dump.

AI and agent clients use controlled API and MCP surfaces to retrieve bounded context. BaseCite does not promise model adoption, answer placement, ranking, or recommendation.

MCPTooling can request scoped context for a known organization and upload.
APIResponses preserve labels such as customer-submitted and unverified.
No raw downloadPublic AI surfaces do not expose private original files.
No bulk exportEnumeration, dataset dump, and embedding export are outside the public surface.

Public docs: /docs/ai-discovery · /docs/developers · /docs/mcp

Company

Private originals stay protected.

The security page keeps the same aurora field, but slows the motion and thickens the protection membrane. The public message is simple: protect private originals, restrict retrieval, record access, and avoid operating-detail disclosure.

No raw downloadLanding pages never offer public original-file retrieval.
No bulk exportThe product is not a dataset export surface.
Tenant boundaryAccess is scoped by organization and credential role.
Audit eventsReads and writes are represented as handling events, not truth judgments.
Resources

Build against the signal layer.

Developer content stays visually continuous with the public site. Schema, API, and MCP elements appear as data particles and context maps, not as a detached black-box docs template.

OpenAPIUse stable docs to understand bounded retrieval contracts.
MCPUse the MCP guide for tool connection shape and limits.
LabelsPreserve customer-submitted, unverified, and not-evaluated labels.
BoundaryDo not infer truth judgment or answer placement from context availability.

Read developer docs View MCP guide

Contact

Start behind the membrane.

Use the portal only for controlled upload flows. Public pages explain the boundary; private originals are not offered as public downloads, bulk exports, dataset dumps, embeddings exports, or chunk exports.

Consent firstScoped accessBounded retrievalCorrection path
Open controlled portalReview boundariesDeveloper resources