Connect the systems teams already use
Email, CRM, files, tickets, collaboration, and manual uploads become reviewable source material before extraction.
Built for Big Berlin Hack
Context your agents can trust before they act.
Context Surgeon turns scattered operational knowledge into a source-grounded context base, then keeps it current as facts change.
153,997
sample records
7
VFS files
Fact Patch
not regenerate
Sonnenallee 44 context run
public sample data ยท ready to inspect
Claims become source-grounded facts
Tenant email
roof leak still active
Old log
roof issue closed
judge-visible proof
Fact ledgerfact_roof_opentenant quote cited
conflictclosed vs open
spanline-level evidence
Built with Big Berlin Hack partner technologies
Google DeepMind / GeminiTavilyPioneer / Fastino
Product
The missing operating layer between business systems and agents.
01
Build a transparent context base
The system produces typed facts, source quotes, conflicts, and a virtual file system that humans and agents can inspect.
Update only what changed
New evidence updates the right context without overwriting human judgment or trusted notes.
Sample-data proof
The supplied Inazuma dataset becomes a context base, not a prompt dump.
The demo compiles CRM, HR, enterprise mail, support, policies, IT tickets, GitHub, social data, and order artifacts into a virtual file system plus graph.
153,997
raw records
12
domains
9
entity types
Generated repository
source-linked VFS for humans and agents
/company/inazuma/context.md
Company Operating Context
Dense executive context for AI agents and humans.
/company/inazuma/static/customers.md
Customers and Revenue
Static customer, product, sales, and order memory.
/company/inazuma/static/employees.md
Employees and Roles
Employee directory, HR relationships, resumes, ownership references.
/company/inazuma/procedures/policies.md
Policy Repository
Procedural company rules from policy PDFs.
/company/inazuma/trajectory/projects.md
Projects and Progress
Trajectory memory from mail, collaboration, GitHub, support, and IT tickets.
Live mode
Connect the systems teams already use.
Teams choose the sources that matter, review new evidence, and approve what becomes part of the shared context base.
Gmail, Outlook
CRM
HubSpot, Salesforce
Files
Google Drive, OneDrive, Dropbox
Work streams
Slack, Teams, Zendesk, Jira, Linear
Live API
Upload evidence without opening the UI.
The API mirrors live mode: list connectors, create connect links, sync systems, upload evidence, and activate scoped ingestion rules.
01
Connect
02
Normalize
03
Patch
POST /api/live/upload
multipart file intake for live mode
curl -X POST https://contextsurgeon.fnctn.io/api/live/upload \
-H 'Authorization: Bearer <FIREBASE_ID_TOKEN>' \
-F 'files=@owner-email.txt'GET/api/live/connectorsConnector catalog
POST/api/live/connectOAuth link
POST/api/live/syncSync systems
POST/api/live/uploadUpload evidence
POST/api/live/rulesIngestion rule
Input
PDF, Word, text, CSV, JSON, Markdown, image, or exported evidence.
Output
Candidate sources, confidence, fact hints, and next actions.
Agent pre-flight check
actionableEvery fact carries a source quote.
Conflicts appear before an agent acts.
Human notes survive generated updates.
The agent plan changes only after context changes.
Trust model
The product shows its work.
Context Surgeon does not ask people to trust a black-box answer. It exposes source evidence, conflict explanations, patch diffs, preserved notes, and before/after plans.