Skip to content
ortivo
Menu

Ortivo user guide

Reviewed and updated 28 October 2026. Start with the steps below, then use the contents to find a feature. Basic Help & Fix works without an AI connection; an optional AI explanation is a separate reviewed request.

Start here

Ortivo lets you ask AI services, compare their answers and work with documents in one workspace. Web and News search and basic Help do not require a personal AI-provider key.

  1. Sign in to Ortivo and open Chat.
  2. For an AI answer, open Connections and connect one provider. An API key is a private access code that lets Ortivo send requests to your provider account. Check that account’s API billing, credits and usage limits; an ordinary chat subscription does not automatically cover API use.
  3. Choose one connected provider and an available model. Ask a short question. Review costs and recipients before sending.
  4. For websites or news without an AI answer, use + → Web search. Choose Web or News and open the source links.

Connect your first provider · Understand usage costs · How your data is handled

Settings: connections, models and costs

Connect only the providers you intend to use. Provider API access, available credits and model permissions must be checked in each provider account.

  1. Open Settings and choose OpenAI, Claude, Gemini or OpenRouter. Open its official API-key guide below.
  2. Create a dedicated key in that provider’s dashboard. Review account access, API billing, model permissions and available spend controls. No local software installation is required.
  3. Paste the key only into the matching password field in Settings, review the storage consent and click Verify & connect. Never paste it in Help or Chat.
  4. Refresh available models, then choose a model you can access. A configured connection or model list does not guarantee a successful generation request.
  5. Start with a short text-only question to one provider when you choose to test. A test is a paid AI request unless the provider account specifically covers it.
  6. Disconnect removes the active saved key from Ortivo; revoke a compromised key separately in the provider dashboard.

Troubleshooting API connections

Open Settings · Open Chat

OpenAI: obtain and connect an API key

No software installation is needed. Create the key on the official provider website, then connect it in Ortivo Settings. Never enter it into Help or Chat.

  1. Sign in to the OpenAI API platform and choose the intended organisation/project. ChatGPT billing is separate from API billing.
  2. Review API billing, available credits, project permissions and model access. Set practical spend controls or alerts; an alert is not necessarily a hard cap.
  3. Open API keys and create a dedicated secret key for Ortivo. Copy it securely when displayed; if lost, create a replacement.
  4. Return to Ortivo Settings. Paste the dedicated key into the OpenAI password field, review the consent and click Verify & connect.
  5. Refresh models and select a model available to your API account. When ready, send one short text-only question to this provider in Chat; provider charges may apply.
  6. If a key is exposed, revoke it on the provider’s API-key page, create a replacement and reconnect it in Ortivo. Disconnecting in Ortivo does not revoke it at the provider.

Troubleshooting OpenAI API keys

Open Settings · Open Chat

Anthropic Claude: obtain and connect an API key

No software installation is needed. Create the key on the official provider website, then connect it in Ortivo Settings. Never enter it into Help or Chat.

  1. Sign in to the Claude Console and choose the correct organisation/workspace. Review the API account’s own billing and access, including any account-specific credits.
  2. Open API keys. Create a dedicated key for your intended workspace and appropriate use; your account role controls which keys you can create.
  3. Check available API credits/billing, spend/rate limits and model access. Consumer plan names alone do not establish what API usage is covered.
  4. Return to Ortivo Settings. Paste the dedicated key into the Anthropic Claude password field, review the consent and click Verify & connect.
  5. Refresh models and select a model available to your API account. When ready, send one short text-only question to this provider in Chat; provider charges may apply.
  6. If a key is exposed, revoke it on the provider’s API-key page, create a replacement and reconnect it in Ortivo. Disconnecting in Ortivo does not revoke it at the provider.

Troubleshooting Claude API keys

Open Settings · Open Chat

Google Gemini: obtain and connect an API key

No software installation is needed. Create the key on the official provider website, then connect it in Ortivo Settings. Never enter it into Help or Chat.

  1. Sign in to Google AI Studio and open Get API key. Select or create the intended Google Cloud project following Google’s current key setup flow.
  2. Create a Gemini API key using the current AI Studio flow. Key type, API restrictions and project access affect whether it will work. Keep the key private.
  3. Review Gemini API billing, model availability, quota, region and free/paid tier conditions for that project. Do not assume a Google consumer subscription covers this API usage.
  4. Return to Ortivo Settings. Paste the dedicated key into the Google Gemini password field, review the consent and click Verify & connect.
  5. Refresh models and select a model available to your API account. When ready, send one short text-only question to this provider in Chat; provider charges may apply.
  6. If a key is exposed, revoke it on the provider’s API-key page, create a replacement and reconnect it in Ortivo. Disconnecting in Ortivo does not revoke it at the provider.

Troubleshooting Gemini API keys

Open Settings · Open Chat

OpenRouter: obtain and connect an API key

No software installation is needed. Create the key on the official provider website, then connect it in Ortivo Settings. Never enter it into Help or Chat.

  1. Sign in to OpenRouter and open Settings → API keys. Create a dedicated key for Ortivo.
  2. Review credits, model pricing and any per-key credit limit. A free model may still have access/rate constraints; do not assume every model is free.
  3. Check your provider privacy/routing settings. Ortivo’s current integration is text only and requests routing without data collection or endpoint retention, which can reduce available models.
  4. Return to Ortivo Settings. Paste the dedicated key into the OpenRouter password field, review the consent and click Verify & connect.
  5. Refresh models and select a model available to your API account. When ready, send one short text-only question to this provider in Chat; provider charges may apply.
  6. If a key is exposed, revoke it on the provider’s API-key page, create a replacement and reconnect it in Ortivo. Disconnecting in Ortivo does not revoke it at the provider.

Troubleshooting OpenRouter API keys

Open Settings · Open Chat

Chat: ask, compare and follow up

Use one AI for a focused answer, or compare independent answers from connected providers. Each selected provider may charge your API account.

  1. Open Chat and choose Single AI or Compare. Select connected providers and an available model for each.
  2. Write the outcome you want, relevant facts, constraints and preferred answer format. Check the recipient line before sending.
  3. Web off is selected for new chats. Choose Web on manually when needed, or explicitly select Auto. Review brief, memory and attachments before each request.
  4. Send once, wait for completion and read each provider’s result and usage separately. Ask a follow-up in the same conversation or start a new chat for a separate task.
  5. Use Review with AI Council to challenge successful answers. Save a useful decision in the notebook; export before leaving an unsaved session.

Troubleshooting AI chat and comparisons

Open Chat · Open Council · Open Settings

More detail and practical checks

Choosing a model

The dropdown is populated from models returned by your connected provider account. Recommended available models appear first, followed by the remaining models. An absent model is not automatically created or made accessible by Ortivo.

Choose a model appropriate to the task and cost you accept. A model appearing in the list does not establish that it supports every file format or every optional tool. Image, PDF and web-search support differ. Provider permissions, billing and availability can also affect requests.

When Web search is switched on, Ortivo retains a compatible selected model or picks the first supported model in each selected provider’s available dropdown. Providers with no compatible option are deselected with an explanation. Gemini is currently deselected because Ortivo has not enabled Gemini search. No additional provider is added automatically.

Turning Web search off retains the current model choices. If you manually choose an unsupported model while search is on, the request is blocked until you select a supported model or turn search off. Model recommendations and support can change in future plugin releases.

Loading indicators and waiting

During ordinary chat and search, the active answer card displays an animated indicator, a provider-specific message and elapsed seconds. After 20 seconds, the message says that Ortivo is still waiting for the provider. This is elapsed time, not an estimate of time remaining.

Selected providers run in parallel, each with its own loading display. Completed answers remain visible if another request fails. Loading wording does not claim that an AI is verifying a source or completing a particular internal step.

Keep the page open while waiting. Avoid repeated clicks, duplicate tabs or immediate resubmission after an error. Ortivo does not automatically retry failed requests. A browser timeout does not prove the provider stopped processing or did not bill the request. Check the provider usage dashboard before trying again. The interface has no guaranteed provider-side cancellation control.

Chat: images, files and dictation

Attachments are shared only with the selected compatible providers after review. Original files are not carried into later turns or Council.

  1. Use the attachment control and wait for reading to finish. Check the displayed supported formats, size and page limits before uploading.
  2. Review each file and the exact recipient/model preview. Remove sensitive pages and unnecessary images before confirming.
  3. Ask a specific question about the supplied file. For long text and decision-focused reading, use the dedicated Document reader.
  4. For image preparation, use the image optimise/enlarge controls where available; inspect the result before attaching. Enlargement cannot restore facts absent from the original.
  5. For dictation, grant microphone permission only when you want to dictate, review the transcript and stop recording before sending.

Troubleshooting files and dictation

Open Chat · Open Documents

More detail and practical checks

Optimise and enlarge images

  1. Select an image within the upload size limit.
  2. Click Optimise image beside it. Processing takes place in your browser.
  3. Check the size-reduction notice. If the image cannot be made smaller, the original is retained.
  4. Click the thumbnail to open Image preview. Keyboard users can focus it and press Enter or Space.
  5. Use Actual size to inspect detail and scroll where necessary. Use Fit to screen to return to the fitted view.
  6. Close with ×, Escape or the backdrop outside the preview.

Optimisation can resize the longest edge to 2,400 pixels and convert the image to JPEG. It can soften small text and replace transparency with white. Inspect diagrams, labels and screenshots before sending. The preview shows the version currently prepared for upload, not a retained original-versus-optimised comparison.

An image already over 10 MB is rejected before this tool can optimise it. Resize it using a trusted tool on your device, then select it again. There is no built-in PDF, Word or general file compression.

Microphone dictation

  1. Click Microphone.
  2. Read the browser-speech-service notice and agree if you want to proceed.
  3. Allow microphone access in your browser when prompted.
  4. Speak your question. Review the words placed in the composer.
  5. Click the microphone control again to stop if it is still listening. Correct names, numbers and punctuation.
  6. Click Send yourself. Dictation does not automatically submit a question.

This feature converts speech into editable text; it is not a real-time voice conversation, audio upload or spoken-answer feature. Your browser’s speech service may process audio externally. Ortivo does not itself record or upload an audio file for this feature.

Recognition stops when you leave the view or page, the page is backgrounded, or the listening time limit is reached. Browser support and device permissions vary. If recognition is unavailable, type your question. If it is blocked, check browser and operating-system microphone permissions. Never dictate API keys or passwords.

Projects: organise work and reuse sources

Projects organise related work, notes and saved sources. A saved project is not proof that every AI request includes its contents.

  1. Open Projects and create or select a project. Give it a recognisable title and concise notes.
  2. Save useful website or document sources to the intended project using the reader’s save controls.
  3. Reopen a saved source in the appropriate reader. Check that it is still current before using it.
  4. Use the visible context controls and request preview to decide what accompanies a task. Include only the context needed.
  5. Keep decisions in the notebook and reusable instructions in My prompts, rather than duplicating full conversations in every note.

Troubleshooting projects

Open Projects · Open Documents · Open Website · Open Decision notebook

Projects: Document reader

Read extracted document text with a clear question or a reading lens. Extraction and an AI answer are different steps.

  1. Open Document reader and choose a supported file. Wait for extraction and inspect the extracted source text.
  2. Choose Audit claims, Find decision hinge or Extract commitments, or write your own question.
  3. Choose a connected provider and available model. Review the exact source text and question in the paid-request preview.
  4. Send after review. Check the answer against the original document, especially tables, figures, scanned pages and commitments.
  5. Save the source to the correct project when useful. Record a decision separately in the notebook.

Troubleshooting Document Reader

Open Documents · Open Projects · Open Settings

My prompts: browse, personalise and save

Browse 48 starter templates across 11 categories, then turn a useful one into a prompt that fits your task.

  1. Open My prompts. Browse categories or search for the outcome you need. Switch to My saved for your own prompts.
  2. Open a template and fill its required context fields. Review the complete prompt preview for unresolved placeholders and sensitive information.
  3. Prepare a Chat draft, or use Prepare Research Agent goal for supported research templates. This fills a draft and does not send an AI request.
  4. In My saved, create, edit or duplicate a reusable prompt. Use {{variable}} placeholders for details that change between tasks.
  5. Search your saved prompts, use pagination and export a backup when useful. Avoid saving API keys or private source documents in templates.

Troubleshooting saved prompts

Open Prompts · Open Chat · Open Research Agent

Settings: personal and project memory

Memory can hold short facts for reuse. Scope and inclusion should be deliberate; do not use it as a password store.

  1. Open Settings and then Memory using the context/settings controls.
  2. Review existing personal/project facts and choose the intended scope before saving a new fact.
  3. Keep each fact short and factual. A full answer may exceed the 600 UTF-8 byte limit for one memory.
  4. Use the visible memory controls to decide what accompanies a Chat request. Inspect the preview and recipient before sending.
  5. Review, edit or remove outdated facts through the memory controls. Help does not automatically read memory contents.

Troubleshooting personal and project memory

Open Settings · Open Chat

Settings: My brief and context

My brief stores reusable context. Attach it deliberately and inspect which providers will receive it.

  1. Open My brief under Settings. Write only the stable context relevant to your work and save it.
  2. In Chat, enable Attach my brief when you want it included in that conversation.
  3. Review the provider/recipient line and other context controls. Turn off attached brief before web search.
  4. For project-specific tasks, supply the relevant source text explicitly rather than assuming all project notes are automatically attached.
  5. Keep secrets and unnecessary personal information out of reusable context.

Troubleshooting your brief

Open Brief · Open Settings · Open Chat

History: saving, reopening and exports

Saving is off by default. Private-session conversations disappear when you leave the page; enabled saved history belongs to your Ortivo account.

  1. Choose the save-history setting before starting a conversation. Changing it after a conversation has begun starts a new conversation after confirmation.
  2. Open History to view saved conversations. Reopen the intended provider’s conversation before following up.
  3. Use export to download your saved conversations and keep the file securely.
  4. Use deletion controls carefully: active storage deletion may be permanent, while encrypted backups follow their own expiry.
  5. Save decisions and research findings in the notebook for a concise record rather than relying on a long transcript.

Troubleshooting saved history and exports

Open History · Open Chat · Open Decision notebook

Costs, token usage and practical limits

Each answer shows the model and provider-reported input/output token counts. Tokens are units of model input and output, not a direct currency amount. These numbers are not an invoice and may not itemise all search, cached-input or other provider charges. Your provider dashboard is the source of truth.

Use one provider for simple tasks, keep briefs focused, ask concise questions and start a new conversation when old context is unnecessary. Check provider usage and spending controls regularly.

Current plugin limits include 100 chat/Council requests per user per UTC day, up to 20 exchanges in a conversation and an output allowance of 2,048 tokens per request. The daily request limit is not a dollar spending cap. Question and request-size limits can also apply. “Output limit reached” means the answer may be incomplete; a new follow-up is another request and may incur another charge.

AI Council: review, challenge and decide

Council reviews successful answers from your latest Chat request. Each review stage is an additional paid request, with a preview of the material being shared.

  1. Ask your question in Chat first. Keep the successful answers you want to examine on this page.
  2. Open AI Council. Choose which AI should review and select an available reviewer model. Only successful answer sources are selectable.
  3. Choose Evidence check, Stress test, Decision hinge or Action plan. Edit the instructions to state the decision you are trying to make.
  4. Check the source selections and exact request preview. Approve only the material and provider you intend to use.
  5. Read the result for unsupported assumptions and practical next steps. Advanced stages may include earlier Council results; check their preview too.
  6. Save the decision, key assumption, smallest useful test and revisit condition in the Decision notebook.

Troubleshooting AI Council

Open Council · Open Chat · Open Decision notebook · Open Settings

Projects: Decision notebook

Keep the decision and its reasoning together so you can revisit it when evidence changes.

  1. Open Decision notebook and add a record, or save from Council or Research Agent.
  2. Record the decision question, evidence, chosen action and uncertainty. Use Add decision hinge & review prompts for a structured starting point.
  3. State the key assumption, smallest useful test and the condition that would make you reconsider.
  4. Save changes. Use search and status filters to find evidence-only, decision and outcome records.
  5. Add the observed outcome later. A status reflects which fields are filled; it does not verify evidence quality.

Troubleshooting Decision Notebook

Open Decision notebook · Open Council · Open Research Agent

Research Agent: brief, plan, research and report

A bounded run can use up to four paid requests: one planning request, two research requests and one report request. That is a request limit, not a spending cap.

  1. Open Research Agent. Use the brief builder to define the decision, audience, scope and desired evidence, or write a focused goal under 2,000 UTF-8 bytes.
  2. Select a connected OpenAI or Claude provider and a search-compatible model. Include exact public HTTPS URLs when a particular site matters.
  3. Review and approve Create research plan: one paid request. Edit the two proposed questions before approving research.
  4. Approve the plan for up to three further requests. Inspect scope and domain warnings; off-scope or missing source URLs may pause the run for your review.
  5. Watch request usage and the evidence ledger. Stop after current request prevents later requests but cannot guarantee cancellation or refunds for the active one.
  6. Keep the page open. Export the report and evidence ledger, or save useful results to the Decision notebook before leaving.

Troubleshooting Research Agent

Open Research Agent · Open Sources · Open Website · Open Decision notebook

Research: Website reader

Read one accessible public page and inspect the extracted text before asking an AI about it.

  1. Open Website reader and enter the exact public HTTPS page URL.
  2. Fetch/read the page and inspect the extract and source address. This does not crawl an entire website or sign into private pages.
  3. Choose a reading lens or write a question, select a connected model and review the complete text before approving an AI request.
  4. Check important claims against the original page. Save useful sources to a project and record decisions separately.

Troubleshooting Website Reader

Open Website · Open Sources · Open Projects

Research: Web sources and You.com

Web sources retrieves search results or public page contents through a separate You.com connection. It is separate from the four chat AI providers.

  1. Open Web sources. Connect You.com in its connection area if retrieval is not configured.
  2. Enter a focused query or exact public URL, then review the retrieval request and any service charges.
  3. Inspect result URLs, dates and contents. Add relevant sources to the selection tray.
  4. Use the available handoff/save controls to work with chosen material. Review any AI request separately; selecting sources alone does not send them to an AI.
  5. Keep the source address and retrieval context so you can verify what evidence was available.

Troubleshooting Web Sources and You.com

Open Sources · Open Website · Open Research Agent

Research: Scheduled research

Scheduled research can create private reports on a recurring schedule after you review and enable it. Future runs can incur provider charges.

  1. Open Scheduled research. Create a paused draft with a clear goal, provider/model and frequency.
  2. Review what will run, who receives the request and expected costs. Check the site’s displayed schedule/timezone.
  3. Enable only when you want recurring requests. Keep the scope small and review provider budget controls.
  4. Read reports and check the last/next run status. Pause or disable the schedule when no longer needed.
  5. Use manual Research Agent when you need to approve each plan and report stage.

Troubleshooting scheduled research

Open Scheduled · Open Research Agent · Open Settings

Troubleshooting: guided checks and safe recovery

Check the exact failing feature, the step that failed and whether any request may already have reached the provider. No automatic paid test or retry is run by Help.

  1. Use Check workspace for read-only connection configuration and visible UI checks. A configured connection is not a live health test.
  2. If the session expired, copy unsaved drafts and export results, then sign in again or reload.
  3. For authentication, quota and model problems, open Settings and follow the matching provider guide. Check the provider dashboard/status.
  4. For file or website problems, inspect extraction before sending to AI. For Council, make sure successful Chat answers exist in this page session.
  5. Copy the redacted support summary if you need further help. Review it before sharing; add a short description of expected versus actual behaviour, without keys or private documents.

Troubleshooting guided recovery

Open Settings · Open Help

Navigation, mobile and accessibility

The adjustable sidebar groups tools into Chat, Projects, Research, Prompts, History, Settings and Help. AI Council is available from Chat and its own URL.

  1. Use the sidebar toggle to expand or collapse navigation. On narrow screens, open the menu and select a section.
  2. Use the section tabs for related tools: Projects includes readers/notebook; Research includes Agent, Website reader, Web sources and Scheduled research.
  3. Bookmark a tool’s own URL. Browser Back and Forward return between visited workspace tools. Chat is at the homepage; /workspace/ redirects there.
  4. Use keyboard Tab and Enter for controls and Escape to close Help. Help retains this page’s conversation until Clear or leaving the page.
  5. If a control is hidden, expand the sidebar or open the feature from this Help directory.

Troubleshooting navigation and mobile access

Open Help · Open Chat

Google Drive import

Google Drive import is available to signed-in Ortivo users. Open Chat, choose +, then Choose from Google Drive. A connection panel shows loading progress. When ready, select Continue to Google to open authorisation. Choose the Google account containing the document you want to import and review the requested permission.

Choose one Google Doc, Sheet, PDF, DOCX, XLSX, TXT, Markdown, CSV, JPEG, PNG or WebP file. Import one file at a time; a message can contain up to three attachments, 10 MB each and 20 MB total. Text and extracted Office content are limited to 500 KB per file. In + → More tools → Google Drive permissions & privacy, choose Preserve page layout (PDF) for Google Docs/Sheets, or Extract text and spreadsheet cells for analysis. PDF export retains exported page layout and images; comments are omitted, and large sheets may span pages. DOCX imports main text only. XLSX imports sheet names, cell addresses, values and formulas; formulas are shown without execution, and charts/images/most formatting are omitted. Legacy DOC/XLS, password-protected Office documents and unsupported image types must be converted first. Review the attachment and select compatible AI providers before Send. Import does not automatically call an AI. Large extracted text can exceed a provider model context limit. Computer uploads support the same PDF, Office, image and text formats.

Disconnect Drive clears the token from this page; it does not revoke the Google grant or remove already imported attachments. Remove the attachment, start New chat or leave the page to clear it. Google Account connections provides full grant revocation. See the privacy policy for details. If Google tools cannot load, use a computer upload while the connection is investigated.

Help & Fix

Read the user guide