Finkkle Spaces
Plugins: extend what a conversation can do.
Plugins are focused tools that give Spaces a predictable capability—research, code analysis, file creation, visualisation, communication, or a connected service—without changing the way you work in a conversation.
Feature reference · Finkkle documentation
What a plugin is
A plugin is a packaged capability with a name, description, category, developer, version, and execution boundary. Spaces uses the conversation to decide when a plugin is relevant, asks for the inputs it needs, runs the capability, and returns a result that can be reviewed in the same thread.
Plugins are not separate chatbots and they do not replace the model. The model plans the request; the plugin performs the specialised operation; Spaces presents the output and keeps the action visible in the conversation.
Plugin lifecycle
- Discover: open the Plugins screen from the Spaces navigation. Browse the featured carousel, category lists, or search by name, description, or category.
- Inspect: open a plugin detail view to read its purpose, capabilities, developer, version, rating, run count, bundle size, and sandbox type.
- Install: select Install. The plugin becomes available to the workspace and its action is added to the capability set Spaces can consider.
- Use: describe the desired outcome in a conversation. Mention the plugin by name when you want deterministic routing, and provide the files, links, or account connection it needs.
- Review: inspect the returned artifact, citations, file, message draft, or action log. Ask Spaces to revise the result before sharing or sending it.
- Manage: use the plugin action menu to uninstall a package, view terms, or disable updates. Removing a plugin does not delete the conversation record that referenced it.
How plugin execution works
Every invocation follows the same shape: input collection, permission check, isolated execution, result validation, and conversation hand-off. A plugin receives only the context required for the requested operation. A tool that needs a file receives the selected file; a connected service receives the account scope approved during authorization.
- Input collection: Spaces identifies required text, files, URLs, settings, or connection credentials.
- Permission check: read, write, send, and account actions are distinguished so a draft can be prepared without sending it.
- Sandbox: the detail view identifies the execution boundary, such as AI Sandbox, Web Browser, Command VM, or Document tool.
- Result validation: outputs are checked for a usable response, supported file format, and actionable errors.
- Audit trail: tool output, status, and relevant action details remain reviewable in the conversation or run history.
Choosing a plugin
Choose the narrowest tool that matches the job. Use Deep Research for multi-source investigation, Code Explainer for understanding code, Terminal Executor for reproducible scripts, and a connected service only when the work must reach that service. For a one-off question, ordinary Chat may be faster and require less context.
Built-in catalog
Featured
- Computer Use
- Controls desktop environments through vision-guided navigation, mouse movement, keyboard input, and a consented action log.
- Chrome Sandbox
- Loads JavaScript pages in an isolated browser, reads visible text or DOM content, handles standard overlays, and performs approved web actions.
Data & Analytics
- Deep Research
- Runs multi-step web research, checks sources, synthesises findings, and produces structured reports with tables and charts.
- D3 Visualizer
- Turns structured data into responsive SVG charts, flow diagrams, scatter plots, networks, and heatmaps with exportable visuals.
- SQL Grounding
- Explains relational schemas, proposes analytical SQL, checks indexes, and protects data-changing operations with query safety controls.
Dev Tools
- Code Explainer
- Explains algorithms and syntax, diagnoses logic errors, suggests refactors, and works across JavaScript, TypeScript, Python, C++, and Go.
- GitHub Integration
- Links repository trees, monitors branches and pull requests, drafts issues or commits, and supports code-review workflows with approved authentication.
- Terminal Executor
- Runs bash, Python, Node, git, curl, jq, and related utilities in an ephemeral Linux command environment with time and resource limits.
Productivity
- Workspace Sync
- Drafts Gmail messages, checks calendar conflicts, creates reminders, schedules meetings, and searches connected Drive content through OAuth.
- Notion Sync
- Searches pages and databases, appends summaries, creates structured pages, and updates task records through an approved Notion integration.
Document tools
- Spreadsheets
- Creates and edits grids, imports CSV/TSV and workbook data, evaluates formulas, applies cell styles, and exports spreadsheet files.
- Presentations
- Builds slide outlines and layouts, inserts charts and images, applies presentation themes, and exports PDF or editable PowerPoint packages.
Communication
- Slack Bridge
- Summarises channel activity, drafts rich messages, dispatches approved notifications, and supports Slack formatting and webhook flows.
- Discord Webhook
- Creates embeds, sends developer alerts, pipes selected logs, and queues webhook requests when rate limits apply.
Education
- Study Mode
- Creates flashcards, adaptive quizzes, revision checklists, spaced-repetition loops, and printable study decks from conversations or files.
- Quiz Generator
- Transforms supplied material into question sets with answer keys, difficulty levels, explanations, and practice-ready formats.
- LaTeX Parser
- Renders equations, matrices, calculus, and scientific notation into responsive KaTeX or exportable vector layouts.
Connected plugins and approvals
Connected plugins use an authorization step before they can access an external account. Grant the smallest useful scope. Draft actions should remain drafts until you approve them; sending a message, creating an event, writing a page, or changing a repository should be treated as an external action.
When a plugin supports observe-only or draft-only operation, use that policy for review-heavy workflows. Revoke the connection from the provider when access is no longer needed.
Custom plugins
The Personal segment is designed for custom plugins. A custom plugin can wrap an API endpoint, webhook listener, or local script behind a declared schema. Define its inputs and outputs, describe the authentication boundary, and test it with non-sensitive sample data before adding it to a production workflow.
- Describe the operation in plain language.
- Declare required inputs, optional inputs, output formats, and failure responses.
- Choose the smallest network, file, and account permissions.
- Run it in the builder sandbox with representative test cases.
- Publish only after the result and action log are understandable to another person.
Writing effective plugin prompts
State the operation, source of truth, constraints, and desired output. For example: “Use SQL Grounding to inspect the attached schema. Explain the relationship between orders and refunds, then draft a read-only query for monthly refund rate. Do not modify data.”
Failure modes
- Missing input: attach the required file or provide the URL, identifier, date range, or output format.
- Permission denied: reconnect the provider or reduce the request to a permitted read/draft operation.
- Unsupported format: convert the source to a supported file type and retry.
- Timeout: narrow the scope, split the job, or ask the plugin for an intermediate result.
- Unexpected output: keep the run in the conversation, describe the mismatch, and request a corrected pass.