Everything you need to get the most out of Quiddy — from quick-start guides to in-depth references.
Guides for SEEK personas, SPECS controls, Knowledge Vault, memory, and all other features.
Read docs →Answers to the most common questions about setup, API keys, models, privacy, and data.
Read FAQ →What's new in each release — features, fixes, and improvements across every version.
View changelog →Join the Quiddy Discord to share personas, ask questions, and discuss how you use the app.
Join Discord →Report bugs, ask for help, or request features. We respond to every message.
Get support →What's planned — mobile apps, team features, and what's coming next.
See roadmap →Download the .dmg from the Download page. Open it and drag Quiddy to /Applications. On first launch, macOS may show a Gatekeeper prompt — right-click the app icon and choose Open to proceed.
Quiddy opens a guided setup on first launch. You'll be asked for your name, a short description of what you work on, and a starting persona. This takes under two minutes and bootstraps your AI context immediately.
Open Settings → API Providers. Add your key from Anthropic (sk-ant-...), Google AI Studio, or OpenAI. Keys are stored in the macOS Keychain — they never appear in the database. If you're on Apple's internal network, Floodgate is configured automatically.
SEEK stands for Sensitivity · Exploration · Expressiveness · Knowledge. Each dimension is independently configurable with five preset options plus a free-form custom field, and three intensity levels.
Open the Personas panel from the sidebar footer. Click + New persona. Choose SEEK mode, configure the four dimensions, and click Generate PROMPTED Configuration. The background model assembles a full system prompt and five conversation starters tailored to your configuration.
Each persona can have up to six short conversation starters — prompt chips that appear below the input when the persona is active. Click any chip to pre-fill the message box. SEEK generation suggests five automatically; you can edit or add more in the Execution tab of SPECS.
Each SEEK dimension has three intensities. For Knowledge, these are: none (no prefix), Professional, Expert. For the others: Slightly, Normal, Very. A Very Analytical exploration style is noticeably more rigorous than Slightly Analytical.
SPECS (Semantics · Probability · Execution · Constraints · Security) gives fine-grained control over every aspect of model behaviour, independent of the SEEK-generated system prompt.
Set Tone (Objective, Supportive, Socratic, Authoritative), Style (Analytical, Action-Oriented, Explanatory), and Role (Technical Specialist, Creative Writer). Each maps to a specific system prompt injection.
Controls temperature (0 = deterministic, 1 = maximum variation), Top-P nucleus sampling, Repetition Penalty, and Topic Novelty Bias. Hover any slider for a precise description of what it does.
Output Length: Concise (<250 words), Standard (250–600 words), or Comprehensive (no limit). Output Format: Auto, Plain Text, Markdown, or JSON.
Context Memory (how many prior exchanges to include), Knowledge Base (vault/chat folder links), Web Search (Gemini only), and Conversation Starters.
Context Memory scope, Data Privacy (regex-based PII redaction on typed messages), and Content Boundaries (Standard or Strict — Strict refuses medical, legal, and financial advice).
The vault lets you attach documents to personas so they're injected as context on every message. Supported types: PDF (10 MB), text/code/JSON/Markdown (2 MB). Images are not accepted in the vault — they're text-only injection.
In the persona editor, open the Knowledge section and select vault folders or individual documents. On each message, up to 12,000 characters per document are injected into the system prompt.
Vault docs are always-on context — injected on every message for a persona. Chat attachments are per-message: drag or paste a file into the chat input and it's included only in that message. Images and PDFs attached to chat messages are sent directly to the model using vision/document APIs.
After each conversation, Quiddy extracts memorable facts about you using the background LLM model. These facts accumulate over time and are injected into the system prompt as "About You" context.
Extraction is queued and debounced — not run after every single message. After a 30-second quiet period, up to five conversations are processed in a single batch. Facts are deduplicated before insertion. Conversations with fewer than 4 messages or very short average message length are skipped.
When at least two new memories are added, the background model synthesises them into a 1–2 sentence "You are..." persona description. This description is injected at the top of every system prompt.
Open Behavior → Memories to view, edit, or delete individual facts. Use the Rebuild button to regenerate the persona description from current facts.
Anthropic (Claude 4 Opus, Sonnet, Haiku), Google (Gemini 2.5 Pro and Flash), OpenAI (GPT-4o, o-series), and any OpenAI-compatible endpoint including Ollama. Floodgate is available to Apple network users.
In Settings, you can set separate models for Chat (the interactive model you converse with) and Background (the cheaper model used for memory extraction, title generation, and persona rebuild). A typical configuration: Claude Sonnet for chat, Claude Haiku for background.
When using a Gemini model, enable Web Search in SPECS → Execution to add Google Search grounding. The model can retrieve current information before responding.
Press ⌘K to open full-text search. Quiddy uses SQLite FTS5 to search all message content. Results show which conversation matched and highlight the matching text.
Use the filter bar to narrow by: Folder, Tag, Model, Message Role (user or assistant), and Date range. Toggle "Titles only" to search conversation titles instead of message content — useful when you remember what you called a conversation.
All data is stored in ~/Library/Application Support/Quiddy/. Open Settings → Privacy & Data to see exact counts, your data directory path, and options to export or delete everything.
Click Export all my data to download a ZIP containing your SQLite database, vault documents, and attachments. Suitable for backup or migration.
Click Delete all my data and confirm by typing the phrase. All conversations, memories, vault documents, and attachments are deleted. Settings and API keys are preserved so the app still functions.
Share personas, discuss workflows, and get help from other Quiddy users.
Found something broken? Use the built-in Report Issue button (bottom-left of the sidebar) to send a bug report with relevant logs attached automatically.
Email supportHave an idea? We read every feature request. Email us or post in the Discord #feature-requests channel with as much context as you can.
hello@quiddy.appEvaluating Quiddy for your organisation? We can provide security documentation, privacy impact assessments, and volume licensing information.
enterprise@quiddy.appWriting about Quiddy? Find screenshots, logos, and product descriptions in the press kit. Reach out for interviews or early access.
press@quiddy.app