# Renbase

> Governed context for AI agents: one versioned, human-approved corpus of documents and business definitions, served over MCP and REST with typed citations. When the corpus cannot support an answer, the system abstains.

Renbase is built by Renaiss AI. Key facts:

- Agents fail in production on missing business context, not on the model. Nobody told them whether "revenue" means ARR or recognized revenue here, which of four look-alike tables is authoritative, or that USCAN deals after 2025 live in Affinity. That knowledge exists in the company, scattered across Slack, PDFs and people's heads, never in a governed form.
- The corpus holds two kinds of content in one hybrid index: ingested documents (PDF, DOCX, PPTX, XLSX, HTML, Markdown, plain text) and first-class business definitions of four kinds (metrics, entities, rules and glossary terms), each with scope, version, provenance and approval.
- The answer contract has three non-negotiable rules. Sources always: every claim is anchored to an approved definition or an ingested document, and version, approver and last verification travel with the citation. Human approval: machine-generated or imported definitions land as drafts and stay invisible to the answer engine until a team member approves them. Silence over invention: missing, expired or contradictory content produces an explicit abstention, or surfaces both conflicting sources with provenance, rather than inventing or silently choosing.
- Capture, govern, serve. Documents are uploaded through the workspace, the REST API or the SDK; dbt artifacts import as draft definitions and a client-side CLI can propose entities from a warehouse schema without sending credentials or row data. Edits create immutable new versions, feedback on a bad answer can be promoted into a rule, and stale sources are flagged.
- Agents connect over MCP (Streamable HTTP) or REST with per-organization API keys. Tools: get_definition (exact, deterministic resolution by name or alias — free), list_entities (free), search_context (hybrid retrieval — 1 credit), expand_source (unfold a citation into full context — free), ask (complete cited answers — 1 credit). The tools enforce the same answer contract as the human workspace.
- Available today: full REST API, official SDK, MCP endpoint, workspace UI, dbt artifact import and client-side warehouse schema introspection. Coming soon: native connectors (Google Drive, SharePoint, Salesforce, Confluence, Notion and others) and direct connections to databases and data warehouses. Everything they produce will land as drafts requiring human approval.
- Against managed RAG (Bedrock Knowledge Bases, Vertex AI RAG, Azure AI Search): managed RAG indexes documents and returns relevant passages; Renbase treats the answer as a contract, cited or abstained, with human-approved definitions as first-class objects that resolve exactly by name rather than by similarity.
- Neutral by design: Renbase serves context, the agent acts. It does not run SQL against the customer's warehouse or hold its credentials.
- Two deployments. Multi-tenant SaaS: isolated corpus, members, API keys and usage per organization, sign-in by one-time email code, credits consumed per question, no seats and no subscription. Enterprise BYOC: the full stack runs inside your cloud or on-premises and the corpus, questions, answers and usage never leave your infrastructure.

## Docs

- [What is a context layer?](https://renbase.ai/context-layer/): the category explained, with a map of what Renbase implements (including what is not built yet)
- [Glossary](https://renbase.ai/glossary/): tribal knowledge, context drift, abstention, typed citation, provenance, answer contract, MCP
- [Security model](https://renbase.ai/security/): isolation, credentials, what data travels, BYOC
- [Documentation](https://docs.renbase.ai/): architecture, API reference and guides
- [Comparisons](https://docs.renbase.ai/compare/): vs. managed RAG, semantic layers, Snowflake Cortex Analyst, dbt MCP server and Databricks Genie, including when they are the better choice

## Access

- [Request access](https://renbase.ai/contact/): the product is not open to public sign-up yet. Access to Renbase Studio — the workspace for answers, glossary governance and agent keys — is granted after contact.

## Optional

- [llms-full.txt](https://renbase.ai/llms-full.txt): extended product description for LLMs, including MCP tool descriptions, the corpus data model and the FAQ