Agent pipeline integration
Knowledge retrieval participates as a named step in multi-agent workflows — invoked by agents when grounded context is needed, with results flowing directly into the next action.
The Knowledge Base That
Powers Your Agents
Your organization's knowledge — policies, playbooks, docs, data — is only useful when your agents can reach it. Jarvis Knowledge Base surfaces content from 12 supported source types as 1 governed, agent-callable tool that integrates directly into your agentic workflows through the Agent Gateway.
Core capabilities
Built for agents, not just chat
Standard knowledge bases surface answers. Jarvis Knowledge Base makes your enterprise content callable — a live tool that agents invoke, reason over, and act on inside governed workflows.
MCP retrieval tool
Registers as an MCP server endpoint and is auto-discovered by the Agent Gateway. Any connected copilot — Claude, GitHub Copilot, VS Code, Cursor — can invoke it without additional wiring.
Multi-source intelligence
Ingests from cloud storage, wikis, databases, and file uploads. Every source is smart-chunked, vector-embedded, and metadata-tagged into a single governed knowledge layer.
Document citations
Every retrieval response includes passage-level citations traceable to the source document — giving agents verifiable evidence and compliance teams a complete audit record.
Embedded guardrails
PII filtering, topic detection, and message logging are built into the retrieval layer. RBAC scopes enforce which agents and teams can access which knowledge domains.
Automatic sync
Configurable sync schedules keep every knowledge domain current. Deletion safeguards prevent accidental mass-removal, and ACL updates propagate immediately without re-indexing.
01 — MCP Retrieval Tool
Discoverable by any agent, without manual wiring
Once configured, Jarvis Knowledge Base registers itself as a named MCP tool in the Jarvis Registry. The Agent Gateway's auto-discovery sweep surfaces it alongside every other registered MCP server — so any copilot or agent connected to the gateway can invoke the knowledge base immediately.
No per-client configuration. No duplicated access management. The knowledge base tool is available wherever your agents are.
- Registers as
jarvis-knowledge-baseMCP tool on setup - Agent Gateway auto-discovery on every startup sweep
- One authenticated endpoint — every connected copilot
- Tool schema and capabilities auto-generated from KB metadata
02 — Agent Pipeline Integration
Knowledge retrieval as a workflow step, not a sidebar
Agents invoke the knowledge base inside a workflow the same way they invoke any other registered tool — as a step, with inputs and outputs that flow into the next action. One task can query the knowledge base multiple times, across different scopes, with each retrieval informing what happens next.
Every invocation is identity-bound, audit-logged, and traceable to a unique audit ID — so compliance teams have a complete record of what the agent retrieved and why it responded the way it did.
- Multiple KB invocations within a single agent task
- Retrieved context flows directly into downstream workflow steps
- Document citations included in every retrieval response
- Unique audit ID per invocation — traceable to source and user identity
03 — Access Control
Every KB query is identity-bound and domain-scoped
RBAC scopes and ACL policies from the Jarvis Governed AI layer determine which agents, teams, and users can access which knowledge domains — enforced at query time, not just at login. Legal, compliance, sales, and engineering each retrieve only from domains they're authorized for, even when sharing the same underlying index.
- Per-domain RBAC scopes — one index, multiple access tiers
- ACL policies inherited from the Governed AI layer — no duplicate config
- Every retrieval logged with invoking identity, domain, and citation
- Unauthorized queries blocked before retrieval, not after
KB access policy · domain isolation
Knowledge domains
🔒legal-policieslegal-team only
🔒compliance-docscompliance / legal
🌐sales-playbookssales-all
🌐product-docsall-staff
Last retrieval decision
agent: compliance-checker
domain: legal-policies
result: authorized · logged
Connectors
Connect every knowledge source your organization already uses
Jarvis Knowledge Base ships with production-ready connectors for the platforms where enterprise knowledge lives. Each connection becomes a governed knowledge domain — agent-callable through a single endpoint.
Amazon S3
PDFs, DOCX, images, markdown. Cross-account IAM support.
Google Drive
Docs, Sheets, Slides — full Drive hierarchy with OAuth sync.
OneDrive
Personal and shared folders across your Microsoft 365 tenant.
SharePoint
Sites, document libraries, and team drives from Microsoft 365.
Confluence
Wikis, runbooks, and policies from Atlassian Cloud or Data Center.
Notion
Pages and databases from any Notion workspace, synced on schedule.
Web Crawler
Crawl and index public or internal web content on a configurable schedule.
Amazon OpenSearch
Existing search index as a hybrid vector + lexical knowledge source.
PostgreSQL
Tables, views, and document columns as a queryable knowledge domain.
MySQL
Relational tables and BLOB columns indexed as agent-callable knowledge.
Microsoft SQL Server
MSSQL tables — on-premises or Azure SQL, via secure credentials.
File Upload
PDF, DOCX, images up to 500 MB — OCR, chunking, and metadata extraction.
Agent Gateway
Explore the Agent Gateway →The knowledge base becomes powerful when agents can find it
The Jarvis Agent Gateway auto-discovers every registered MCP server — including the Knowledge Base — and routes agent requests to the right tool through a single authenticated endpoint. No per-tool wiring, no duplicated access management.
Ready to make your enterprise knowledge agent-ready?
See how Jarvis Knowledge Base registers as a governed MCP tool and integrates directly into your agentic workflows.
