Documentation Index
Fetch the complete documentation index at: https://docs.snorbe.deskrex.ai/llms.txt
Use this file to discover all available pages before exploring further.
April 2026
Agent-to-Agent mention chains (April 18)
You can now @mention multiple Agents in chat to delegate and chain runs across them. What changed:- Address a specific Agent directly with
@agent-name(1:1 delegation) - Mention two or more Agents and Snorbe reads your intent to pick the flow automatically
- “Give me each of your opinions” → runs in parallel
- “A drafts, then B reviews” → runs as a sequential chain
- An Agent that was addressed can itself
@-mention another Agent mid-reply, so multi-step hand-offs emerge naturally - Chain depth is capped (default 10 hops) for safety. The API exposes
maxChainSteps(1–50) for tuning
- You want a researcher → writer → reviewer pipeline without manual orchestration
- You want independent opinions from Agents with different personas, in parallel
Interactive API Reference (Scalar UI) (April 17)
A Scalar-powered UI for browsing, searching, and executing every REST API endpoint from your browser is now live. What changed:- Scalar UI is available at
https://app.snorbe.deskrex.ai/api/v1/docs - Browse parameters, responses, and code samples (curl / JS / Python, etc.) inline for every endpoint
- Send real requests with your own API key via the “Try it out” button
- Conforms to OpenAPI 3.1 — auto-rendered from the underlying
openapi.json
- Getting an overview of the API before inspecting paths with jq
- Quickly testing endpoint behavior before writing a Slack bot or internal script
Claude Opus 4.7 Support (April 17)
Anthropic’s latest model, Claude Opus 4.7, has joined the available lineup. What changed:- Claude Opus 4.7 is selectable from the model menu
- Supports Adaptive Thinking and a 1M-token long context
- When wired into Snorbe Quality, it’s used preferentially on reasoning-heavy steps
- Running complex inference over large volumes of internal documents
- Tackling tough research themes where existing models feel too thin
Skill Secret Status (April 17)
You can now see at a glance whether the secrets (API keys, etc.) required by a custom skill are configured. What changed:- Skill list and detail screens display each required secret and its configuration status
- Missing secrets are flagged as warnings, with a direct link to the registration screen
- Before using a shared skill, you want to check what’s missing in your own environment
Public REST API for External Integrations (April 16)
Snorbe’s previously internal API is now exposed as a REST API conforming to the OpenAPI spec — so external apps and scripts can drive research agents programmatically. What changed:- REST endpoints served under
https://app.snorbe.deskrex.ai/api/v1/ openapi.jsonavailable as a machine-readable spec- Coverage includes agent execution, chat lists, graph entities/edges, and agent-run status
- SSE streaming for real-time subscription to research progress events
- All endpoints support API key authentication
- Calling Snorbe research from Slack bots, internal tools, or batch scripts
- Streaming long-running research progress into an external dashboard
Orphaned Agent Run Recovery (April 16)
Added a safety net that automatically detects and recovers agent runs left hanging for any reason. What changed:- “Orphaned” runs stuck in running state are detected based on event activity
- Detected runs are automatically resumed or cleanly terminated
- A composite index on
agent_runs(status, updatedAt)speeds up detection
- Confidently delegating long scheduled research runs
Image Lightbox and Body Links on Event Cards (April 16)
Research event cards now display body links alongside images, and images open in a lightbox on click.Three Patent Search Skills (JPO / EPO OPS / Google Patents) (April 15)
A set of official skills covering patent data was added. New official skills:- JPO (Japan Patent Office): Search Japanese patents and utility models
- EPO OPS (European Patent Office): Retrieve European and international patent bibliographic data
- Google Patents: Cross-search patents worldwide
- Sweeping prior-art and competitive filings for a technology theme in one pass
Image Editing and Data Analysis Official Skills (April 15)
Official skills for handling images and data during research were added. The agent runs code in an E2B sandbox and produces artifacts. New official skills:- Image Editing: Crop, resize, and transform images inside the agent
- Data Analysis and Visualization: Analyze data such as CSV and generate charts
- Visualizing collected data for embedding in reports
- Trimming or retouching referenced images before attaching them to research output
Visual Artifact Previews (April 15)
HTML, SVG, and other visual artifacts generated by skills are now shown as PNG previews in the chat. What changed:- Skill completion cards include a preview image
- Generated artifact files can be downloaded from the output files list
Image Gallery in Agent Run Details (April 15)
Images collected during research are shown as a gallery in the agent-run detail panel. Below answers in chat, an image gallery strip is displayed as well. When it’s useful:- Browsing images collected across multiple sources in one view
Enhanced Paper Search Skills (April 15)
The paper search skills (arXiv / PubMed / Semantic Scholar) now use more sophisticated strategies. What changed:- Multi-angle strategy to search a theme from several angles
- Citation traversal to follow references and citations
- Filters for publication year, journal, and more
- Collecting related research comprehensively and following citation networks around key papers
Per-Skill Model Override (April 15)
Custom skills can now pin the model used during execution via amodel field in SKILL.md frontmatter.
What changed:
- Each skill can declare “this skill uses Claude Opus 4.7,” for example
- When unspecified, the skill continues to follow the session’s model setting
- Pinning a high-precision model to a specialized skill that needs it
Information-Gathering Tools in Viewpoint Matrix (April 15)
The viewpoint matrix generation loop can now call information-gathering tools directly — web search, X search, browsing, skills, recall, and agent memory. When it’s useful:- Filling in empty matrix cells by kicking off additional research on the fly
Catalog Recall (New Recall Mode) (April 14)
A new recall mode for browsing an existing knowledge graph “like a catalog” was added. What changed:- Combines simple keyword matching with filter search and strategy extraction
- When results are abundant, presents a continuation hint suggesting where to dig next
- Recall event cards visualize progress and results
- Taking inventory of knowledge buried in an existing graph
- Panoramic exploration like “show me all entities in this category”
Tool Scheduling Metadata (April 13)
The UI now surfaces metadata about how the agent plans to run multiple tools in parallel or sequence. What changed:- Each tool call carries an execution order (
executionOrder) - Timeline and detail panels show which tools run in what order
- Sanity-checking the agent’s plan in real time
- Seeing which tools are executing concurrently when running in parallel
Click-to-Edit Agent Fields (April 13)
Agent name, description, and similar fields can now be edited inline by clicking on them. A note about auto-refreshing memory is also displayed. When it’s useful:- Tweaking the role of a long-running agent mid-stream
Skill Tool Guardrails (April 13)
Added guardrails that raise the safety bar when custom skills are executed. They act as safety nets against unexpected behavior.Timeline Scroll Fix (April 13)
Fixed an issue where wheel scrolling inside the chat timeline chained up to the parent page.Default Agent Auto-Creation on Workspace Creation (April 12)
Creating a new workspace now automatically provisions a ready-to-use default agent. What changed:- A default agent named
snorbe_ai_botis created alongside the workspace - No need to pick an agent before starting your first chat
- Getting started instantly on your first Snorbe touch, without any setup
Unauthenticated / No-Workspace Graph Hero (April 12)
The graph screen now shows an introductory visual (GraphEmptyStateHero) when you’re not signed in or don’t yet have a workspace. When it’s useful:- Helping first-time visitors following a shared Snorbe link grasp the product intuitively
Agent Name Display (April 11)
You can now see at a glance which agent is active in the chat, right as it starts responding. What changed:- An agent name chip appears at the top of each response
- It shows instantly when streaming begins — no waiting
- When you’re using multiple agents and want to know immediately which one answered
Agent Memory (April 11)
Agents now retain memory from previous research sessions and use it to provide more contextual responses. What changed:- Agents accumulate key insights from research and store them as memory
- In follow-up sessions, the agent references that memory to give more informed, connected answers
- When you’re doing ongoing research on the same topic and don’t want to re-explain context every time
- When building up knowledge incrementally over a long-running project
Viewpoint Matrix: Continue Mode (April 10)
You can now add new rows or data to an existing viewpoint matrix without starting from scratch. What changed:- Asking the agent to “add to the matrix” or “extend the table” updates the existing one instead of creating a new one
- Research results can be accumulated in the same matrix over time
- When you want to expand a matrix you built in an earlier session with new findings
X Search Now Shows a Summary (April 10)
X (formerly Twitter) search results now include a summary of the discussion, so you can understand the overall conversation at a glance rather than reading individual posts. What changed:- A summary of key points from relevant posts appears at the top of search results
- Individual posts are still available below
- When you want a quick sense of what people are saying about a topic
Skill Execution Shows Referenced Sources (April 10)
When a custom skill collects information from the web, the sources it referenced are now displayed in the chat. What changed:- Web pages and documents referenced during skill execution appear as source links in real time
- When you want to verify where the skill’s output came from
Improved White Space Detection (Knowledge Gap) (April 9)
The AI now more intelligently analyzes unexplored areas in your knowledge graph and suggests where to research next. What changed:- The AI evaluates overall research coverage in the graph and identifies under-explored perspectives
- These are surfaced as knowledge gaps, with suggestions for next steps
- After a round of research, when you want to know if there are areas you’ve overlooked
Relation Labels on Graph Edges (April 8)
Lines between nodes in the knowledge graph now display relationship labels in both 2D and 3D modes. What changed:- Edges now show labels like “competes with,” “cites,” or “related to”
- The entity detail panel also shows relationship descriptions
- When you want to understand exactly how two pieces of information are connected
3D Animation for Empty Graph (April 8)
A 3D animation now appears on the graph screen when you haven’t started any research yet. When it’s useful:- Makes it easier to know you’re in the right place and invites you to start your first research session
Official Skills Added: CB Insights, PitchBook, Statista (April 1)
Three new official skills are available for market research, investment data, and statistical analysis. New official skills:- CB Insights: Research startups, investment trends, and market maps
- PitchBook: Look up venture capital, M&A, and company valuation data
- Statista: Retrieve market size data, statistics, and industry reports
- When you need specialized data from professional databases in one research session
Viewpoint Matrix: Edit Mode (April 3)
You can now directly edit a viewpoint matrix generated by the AI — fixing, adding, or removing content manually. What changed:- Click any cell to edit its contents directly
- Add or remove rows and columns
- Edit the matrix title freely
- When you want to correct or supplement what the AI generated
- When you want to add your own knowledge directly to the table
Viewpoint Matrix: Improved CSV Import (April 5)
When importing a matrix from CSV, irrelevant rows are automatically filtered out for cleaner, more accurate data.March 2026 (continued)
PDF Export (March 31)
You can now export chat responses directly as a PDF file. What changed:- An “Export as PDF” option is available from the response menu
- The generated PDF includes referenced sources organized as a citation list
- When sharing research results with your team or management
- When you need a printable document for a meeting
Secret Value Preview (March 31)
When registering API keys and other secrets, the first few characters are now shown as a preview. What changed:- A masked preview of the value appears in the secret list and selection dropdown
- When managing multiple secrets and you need to tell them apart at a glance
Install Official Skills (March 30)
Official skills provided by Snorbe can now be installed directly into your workspace. What changed:- Official skills are listed in the Skill Management screen
- One click installs a skill into your environment
- Installed skills can be further customized
- When you want to get started with a ready-made research template without writing one from scratch
Similar Entity Link Highlighting Improved (March 30)
When similar entities are selected, the connecting edges are now highlighted more clearly in both 2D and 3D graph modes. When it’s useful:- When you want to visually trace the connections between entities with similar characteristics
Delete Agent Run (March 29)
Past research sessions (agent runs) can now be deleted directly from the graph’s detail panel. What changed:- A delete button has been added to the agent run detail panel
- A confirmation dialog prevents accidental deletion
- When you want to remove test runs or failed research from your graph
Major Viewpoint Matrix Improvements (March 28–29)
The viewpoint matrix has been significantly improved for easier editing and navigation. What was added / improved:- Drag rows to reorder them
- Select cells to control the full matrix state
- Edit the matrix title inline
- Each cell now links to its referenced source
- Column and row sizes stay consistent between compact and fullscreen views
Custom Skills Released (March 19)
A new “Custom Skills” feature lets you add your own research procedures to the Snorbe agent. What changed:- Define research steps, tools, and output format using a Markdown-based format
- Call skills from chat to run them automatically
- Complex processing including code execution can be automated
- When you want to automate a research workflow you run repeatedly
- When you need research skills tailored to a specialized domain
Dashboard AI Analysis (March 16)
The dashboard can now generate an AI-powered summary of your workspace’s research activity. What changed:- An “Analyze” button fetches an AI summary of your workspace activity
- Set a date range to understand research trends for a specific period
- Private uploaded files are now also reflected in the domain ranking
- When you want to periodically review the state of your workspace
- When you want to share progress with your team
Entity Inline Editing and Merge (March 15)
You can now edit entity information directly on the graph, and merge duplicate entities into one. What changed:- Click an entity’s name or attributes to edit them in place
- Select duplicate entities and merge them into a single node
- Merged entities carry over associated research history and community assignments
- When the same company or concept was captured under slightly different names
- When you want to clean up or correct information in the graph
Entity Comparison: Modal View and Similarity Scores (March 15)
Entity comparison results now appear in a dedicated modal, and similarity scores between compared entities are displayed numerically. What changed:- Results are shown in a clear modal layout
- Each entity pair has a similarity score
- Clicking an entity name zooms directly to that node in the graph
Graph Node Search Dialog (March 15)
A search dialog has been added to find any entity or source in the graph by keyword. It’s accessible from both the chat and graph views. What changed:- Search for any node by keyword and jump to its location in the graph instantly
- Optionally include entities from other workspaces in the search
- When there are too many nodes to scroll through manually
- When you need to quickly locate a specific entity
Graph 2D/3D State Synced to URL (March 15)
The graph display mode (2D or 3D) is now saved in the URL. Share the URL to give someone the same view.March 2026
Private Source Dashboard Redesign (March 14)
The dashboard for managing your uploaded files and URLs has been significantly improved, making it much easier to understand the status of each source at a glance. What changed:- An icon and tooltip now show whether a file is ready to be searched
- A badge appears on sources that have been referenced in chat
- The last update time is shown directly in the list
- Clicking a row expands its details inline, without leaving the page
- When managing many files and you want to quickly check their status
- When you need to know when a specific file was last updated
More File Formats Supported in Preview (March 14)
File preview now supports a wider range of formats. You can view file contents directly from the dashboard or the knowledge graph — no download needed. Supported formats:- PDF (page by page)
- Images (PNG, JPG, etc.)
- CSV (displayed as a table)
- Markdown / Plain text / HTML
- When you want to check what a source file contains without leaving the graph
- When managing many files and need a quick way to see their content
All-Workspace Schedule List (March 13)
You can now see all automation schedules across every workspace you own, in one place from your user settings. What changed:- A workspace badge shows which workspace each schedule belongs to
- All schedules appear in a unified list
- When you manage multiple workspaces and want to see all automations at once
Zoom Buttons Added to Graph Detail Panels (March 13)
A zoom button has been added to entity and AgentRun detail panels, so you can jump directly to the relevant node in the graph with one click. When it’s useful:- When you’re reading details in a panel and want to immediately see where that node is in the graph
Graph 2D/3D Switch Display Fix (March 14)
Fixed an issue where switching between 2D and 3D graph modes left some nodes stuck in a loading state. What was fixed:- After switching modes, some nodes remained frozen in a loading indicator and never resolved
Automation Schedule Bug Fixes (March 14)
Several issues in the schedule feature have been fixed to improve reliability. What was fixed:- Schedules were sometimes displayed at the wrong time due to timezone handling
- Settings toggles were occasionally applied twice unintentionally
February 2026
Fact-Check with One Click (February 27)
You can now ask the AI to verify the accuracy of its responses with a single button click. When it’s useful:- When you want to double-check an important finding before using it
- When a response feels uncertain and you want the AI to confirm it
Send Feedback from Within the App (February 20)
You can now send feedback or bug reports directly from the app interface, without leaving Snorbe. When it’s useful:- When you spot a bug and want to report it right away
- When you have a feature idea and want to send it to the team
Credit System Introduced (February 27)
A credit system has been introduced to help you track and manage your usage. How it works:- Credits are granted daily when you log in
- Your current balance is always visible in the sidebar
- A clear message appears when your balance runs low
- When you want to understand how many research sessions you can run
- When managing usage across a team
Drag & Drop Multiple Files at Once (February 20)
You can now drag and drop multiple files into the chat input area to upload them all at the same time. When it’s useful:- When you want to attach several documents to a research session quickly
Graph Alignment Layout (February 28)
A new layout option automatically arranges graph nodes by category or type, making complex graphs easier to read. When it’s useful:- When your graph is getting crowded and hard to navigate
- When you want to quickly see which nodes belong to which category (person, organization, place, etc.)
Latest AI Models Added (February 28)
The available AI models have been updated to include the latest versions. New and updated models:- Claude 4.6 (Sonnet / Opus): Anthropic’s latest model with extended thinking for complex research
- Gemini 3.1 Pro: Google’s latest large-scale model
Plan Structure Streams in Real Time (February 26)
When generating a research plan, the structure now appears on screen progressively as it’s created, rather than all at once. When it’s useful:- When you want to start reading the plan before it’s fully generated
Detailed Tool Approval Settings (Late February)
You can now configure more granular settings for when the AI should ask for your approval before taking certain actions. When it’s useful:- When you want to automate most tasks but still review specific sensitive actions
Community Boundary Visualization Improved (February 28)
Groups of related information (communities) in the knowledge graph are now displayed with clearer visual boundaries. When it’s useful:- When you want to intuitively see where one cluster of information ends and another begins
Tavily Search Added (February 18)
“Tavily,” a high-precision web search service, is now available as a search option. When it’s useful:- When you need more accurate and relevant web search results
Click a Node to Zoom In (February 18)
Clicking a node in the knowledge graph now moves the camera to zoom in on that area. When it’s useful:- When you want to explore the connections around a specific piece of information
Delete Entities from the Graph (February 17)
You can now delete entities (individual pieces of information) from the knowledge graph. This helps you clean up incorrect or unnecessary information.Preview Files in a Dialog (February 17)
Uploaded files can now be previewed in a popup directly from the chat or from source nodes in the graph.Auto-Update Settings Redesigned (February 17)
The settings card for scheduling automatic entity updates has been refreshed with a cleaner, more intuitive layout.Extract Private Files into the Graph (February 17)
Information from uploaded private files (such as PDFs) can now be extracted directly into the knowledge graph. When it’s useful:- When you want to incorporate internal documents or confidential files into your research
January 2026
Scheduled Automated Research (January 30)
You can now set up automations that run research at a specific time or on a recurring schedule. When it’s useful:- When you want to automatically collect the latest market updates every Monday morning
- When you need regular, repeating research tasks without manual effort
Analytics and API Key Management in the Dashboard (January 30)
The dashboard now includes statistics about your knowledge graph and tools for managing API keys. When it’s useful:- When you want to monitor your workspace’s activity and growth
- When setting up integrations with external tools via the Snorbe API
Reference Information in Chat with @Mentions (January 24)
You can now type@ in the chat to reference specific entities or past research sessions and include them directly in your next request.
When it’s useful:
- When you want to say “@Apple — research this further” and have the AI focus on what it already knows
- When building on a previous research session in a new conversation
Stop Button for Research in Progress (January 10)
A stop button now appears while research is running, so you can cancel it at any time. When it’s useful:- When you realize the research is going in the wrong direction and want to start over
Research Progress Shown in Real Time (January 10)
While research is running, you can now see what the AI is doing at each step as it happens. When it’s useful:- When you want to know how far along the research is
- When you’re curious about how the AI is approaching your question
Browse Settings You Can Adjust (January 27)
The maximum number of steps the AI takes when browsing websites can now be configured from the settings panel.Node Layout Is Now Saved (Mid-January)
The positions of nodes in the knowledge graph are now saved, so the layout you arrange is preserved when you come back. When it’s useful:- When you’ve spent time organizing your graph and don’t want to redo it every visit
Real-Time Browsing Interaction (January 16)
When the AI is browsing the web, you can now give it instructions or answer its questions in real time during the session. When it’s useful:- When you want to redirect the AI to a different link mid-session
- When the AI encounters a login form or other interactive element and needs your help
Entity Detail Panel in the Graph (January 18)
Clicking on a node in the knowledge graph now opens a detail panel showing related research history and source links for that entity. When it’s useful:- When you want to know which research session a particular piece of information came from
Find Similar Entities in the Graph (January 18)
Snorbe can now automatically find and surface entities in the graph that are semantically similar to the one you’re looking at. When it’s useful:- When you want to discover related information scattered across your graph
- When looking for patterns or clusters you might have missed
Copy Options for Chat Responses (January 21)
You can now choose the format when copying a chat response — for example, copying as Markdown.Graph Label Density Auto-Adjusts (January)
Labels in the knowledge graph now automatically show or hide based on your zoom level, keeping the view clean at any scale.For feedback or feature requests, please feel free to contact us.