Step 01
Organize your content
Create folders, upload files, or author documents directly in Opal. Libraries give you additional flexibility for content that doesn't fit neatly into a single folder hierarchy.
Knowledge
One source of truth for your people and your agents.
Knowledge is Opal's centralized knowledge base. Organize, govern, and share the documents and files your organization depends on — so every agent works from the same trusted content your people use.
Knowledge · librarypublished
policy.pdf
ocr · markdown
kickoff.m4a
transcribed
diagram.png
annotated
model-aware routing
native format or text equivalent
agents
grounded
tasks
grounded
threads
grounded
flows
grounded
spaces
grounded
Why most knowledge bases fall short
Your knowledge base was built for humans. Your agents need it too.
Most organizations maintain knowledge in scattered locations — shared drives, wikis, email threads, and local files. When agents enter the picture, teams create a second copy: a separate, often outdated dataset fed to AI tools.
The result is duplication, drift, and agents acting on information that doesn't reflect what your people actually know.
01
Scattered by design
Files live in too many places. When information is spread across systems, no one — human or agent — can rely on a single source.
02
Duplicate maintenance
Feeding AI tools requires copying and converting content. When the original changes, the copy lags behind.
03
Format friction
Raw files aren't always usable. Scanned PDFs, recorded calls, and embedded images are hard to search, hard to share, and even harder for agents to act on.
Knowledge
One place for everything your organization knows.
Knowledge is Opal's built-in knowledge base: a familiar system of files and folders where your team stores, organizes, and governs the content that powers their work — and the work of every agent in your Spaces.
- 01
Files and folders
Organize content the same way you always have. Create folders, upload files from your device, or author new documents directly in Opal.
- 02
Libraries
Group related files and folders in a more flexible, decentralized way. Libraries let you organize content for a specific purpose without restructuring your entire knowledge base.
- 03
Automatic file processing
Upload any file and Opal detects its type automatically. Documents, images, audio, and video are supported — with OCR digitization to markdown, autocorrection, content transformation, image and video annotation, transcription with speaker diarization, and extraction of embedded artifacts into their own files.
- 04
File-type-aware editing
View and edit every file in an interface suited to its format. What you see matches what you uploaded.
- 05
Publishing workflow
Keep content accurate and controlled. Every file has version history, permission settings, and a draft-to-published workflow so agents and teams always work from reviewed, current information.
- 06
Agent access
Add files and folders to Spaces to make them available to agents. Agents can retrieve and ingest content while working in Threads, Tasks, and Flows — and can create or edit documents themselves as part of their work.
Knowledge · pipelinelive
documents
ocr · markdown
audio · video
transcribed
images
annotated
file processing
detect · digitize · transcribe · extract
knowledge
folders · libraries · versions
spaces
flows
tasks
threads
retrieved by agentsgrounded
How it works
Build it once. Put it to work everywhere.
Step 02
Process and prepare
When a file is uploaded, Opal detects its type and surfaces relevant processing options. Digitize scanned PDFs, transcribe recorded calls, annotate images and video, and extract embedded artifacts — turning raw files into content that's searchable, editable, and ready for agents to use.
Step 03
Review and publish
Use the built-in publishing workflow to review drafted or uploaded content before it goes live. Version history tracks every change. Permissions control who can view, edit, or publish.
Step 04
Add to Spaces
Make files and folders available to agents by adding them to a Space. Agents operating in that Space can retrieve and use the content while working in Threads, Tasks, and Flows.
Step 05
Let agents work with it
Agents retrieve content as part of their work — and can create new documents or update existing ones. Opal automatically routes each file in the format best suited to the agent's underlying model, so teams don't need to think about compatibility.
Model-aware knowledge routing
Every agent gets the content in the format it can use.
When an agent retrieves a file, Opal checks whether the agent's underlying model supports that file's format as a native input modality. If it does, Opal passes the file directly. If it doesn't — for example, an audio recording retrieved by a model without native audio support — Opal automatically provides a text-based equivalent, such as the transcription, instead.
No manual conversion. No format mismatches. Agents across your organization can draw on the same knowledge base regardless of which model they run on.
01
Upload once, use everywhere
The same file works for every agent in your organization, regardless of what model they use.
02
No format maintenance
Teams don't need to maintain separate versions of files for different agents or models.
03
Behind-the-scenes optimization
Opal also handles caching and token reduction under the hood, so agents draw on large knowledge bases efficiently without unnecessary overhead.
Why Knowledge works
Trusted content. Everywhere it's needed.
One source of truth
The same content powers human decisions and agent execution. No duplicate copies. No version drift.
Turn raw files into usable knowledge
OCR, transcription, annotation, and artifact extraction make uploaded content readable and actionable — not locked in its original format.
Agents grounded in organizational context
Agents retrieve and act on your organization's real knowledge as they work, and can contribute new documents back to the knowledge base.
Controlled, trustworthy content
Version history, permissions, and a publishing workflow ensure every agent and every team member works from reviewed, current information.
Efficient at scale
Caching and token optimization let agents use large knowledge bases without unnecessary overhead — so performance doesn't degrade as your content grows.
In practice
Knowledge at work across your organization.
Three shapes of governed content, from operating procedures to regulated documentation.
Case 01
Operations
Policy and procedure management
An operations team maintains a set of standard operating procedures, policy documents, and reference spreadsheets. A team member uploads a mix of scanned PDFs and existing documents, uses OCR to digitize the scans into markdown, and publishes the reviewed versions. The folder is added to the team's Space. When a support agent handles a request in a Thread or runs a recurring Task, it retrieves the current, published procedures directly. When a policy changes, the agent drafts an updated document that a team member reviews and publishes.
- Content
- Scanned PDFs · SOPs
- Processing
- OCR to markdown
- Outcome
- Agents cite current policy
Case 02
Proposals
Proposal writing from a living content library
A proposal coordinator maintains a content library of past proposals, boilerplate narratives, past performance write-ups, and staff résumés in Knowledge. When a new RFP arrives, a Flow parses the requirements and agents assemble a first draft from the approved reusable content, tailored to the evaluation criteria. After each submission, agents mine the winning response for new reusable answers, flag stale sections, and draft updated versions for the coordinator to review and publish — so the library keeps improving instead of going out of date.
- Content
- Past proposals · boilerplate · résumés
- Maintenance
- AI-refreshed reusable answers
- Outcome
- First drafts in hours, not weeks
Case 03
Legal and Compliance
Contract and agreement drafting
A legal team maintains contract templates, a clause library, and negotiation playbooks in a governed Library, with permission controls limiting editing to legal. When a business team needs an MSA, NDA, or SOW, an agent drafts it from the approved templates and pulls preferred and fallback clause language for the deal terms, redlining anything outside the playbook for attorney review. When regulations or positions change, legal updates the clause once through the publishing workflow and every subsequent draft uses the new approved language.
- Content
- Templates · clause library · playbooks
- Controls
- Legal-only editing · approved language
- Outcome
- Drafts built from fallback-approved clauses
Security and compliance
Governed content. Controlled access.
01
Role-Based Access Control
Granular permissions at the file, folder, and Library level ensure the right people can access, edit, and publish content.
02
Version History
Every change is tracked. Previous versions are retained and accessible, so content changes are never permanent or unrecoverable.
03
Publishing Workflow
Content goes through review and approval before agents or people can use it. Drafts remain separate from published versions until explicitly approved.
04
Zero Data Retention
Opal does not retain prompts or outputs beyond the active session. Your knowledge base content stays within your control.
05
Audit Log
Every access and edit event is logged. Operations and compliance teams have full visibility into who accessed what and when.
06
Certifications
ISO 27001, ISO 42001, and SOC 2 Type II certifications are underway.
Part of a connected platform
Knowledge powers everything agents do.
01SpacesFiles and folders are added to Spaces to make them available to agents operating in that environment.02AgentsAgents retrieve and act on Knowledge content while working, and can create or edit documents as part of their tasks.03ThreadsAgents can retrieve, create, and edit documents in real-time collaborative conversations with people.04TasksAgents draw on Knowledge while completing recurring automated work, and can author or update documents as an output.05FlowsAgents working through structured multi-step processes retrieve Knowledge at relevant steps and contribute documents back to it.06ModelsOpal routes retrieved files according to each model's supported input modalities, passing native formats or text equivalents automatically.07SkillsSkills define how an agent performs work; Knowledge supplies the organizational information it draws on. The two are complementary.08GuardrailsAccess controls and publishing workflows in Knowledge work alongside Guardrails to ensure agents only act on approved, current content.
Knowledge
Give every agent access to what your organization knows.
Opal's built-in knowledge base keeps your content organized, governed, and ready — so agents and people always work from the same trusted source.