> ## Documentation Index
> Fetch the complete documentation index at: https://docs.oqoqo.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# What's new

> New features, improvements, and bug fixes in Oqoqo.

**November 19, 2026**

**New**

* **[Ask Oqo in GitHub (private beta)](https://docs.oqoqo.ai/operate/ask-oqo-in-github.md):** Mention **@oqoqo-ai** in a comment on an issue, a pull request, or a review line, and Ask Oqo answers in that thread with the repository's code and its Oqoqo project.

**Improved**

* **[Slack accounts in your profile](https://docs.oqoqo.ai/operate/ask-oqo-in-slack.md#manage-slack-accounts):** **Profile → Ask Oqo in Slack** lists each Slack account you connected, and you change its **Default project** right in the list. A new account starts on the project you have open.
* **[Comparisons on the Matrix](https://docs.oqoqo.ai/evaluation/comparisons.md):** **Comparisons** is no longer a separate tab. It is a collapsed section below the **Matrix** grid, and the link in the line above the grid expands it.
* **[Token counts for GitHub Copilot runs](https://docs.oqoqo.ai/evaluation/run-profile.md#time):** GitHub Copilot runs show their token counts in the run profile and in the **Matrix** **Tokens** metric.

**November 18, 2026**

**New**

* **[Eval skills, web pages, and chat files in Ask Oqo](https://docs.oqoqo.ai/operate/ask-oqo.md#what-ask-oqo-can-do):** Ask Oqo uses the Oqoqo Plugin's eval skills across the dashboard, Slack, and your editor to build eval sets, launch experiments, and read results. It can also read public web pages, run commands, and write and upload files from a private workspace for the chat.

**Improved**

* **[Waiting in Ask Oqo](https://docs.oqoqo.ai/operate/ask-oqo.md#while-ask-oqo-works):** Ask Oqo to wait for an experiment, its Evals, or its Insights, and it tells you the result when the work finishes. You can keep chatting or close the page while it waits.
* **[Ask Oqo chat status and recent chats](https://docs.oqoqo.ai/operate/ask-oqo.md#chat-history):** Chat history and the sidebar show whether each chat is working, waiting for you, completed, stopped, or failed. Filter chat history by **Status**, and open your three latest chats from the sidebar or **Recent chats** on **Home**.
* **[Delete many chats in Ask Oqo](https://docs.oqoqo.ai/operate/ask-oqo.md#chat-history):** Chat history has an actions menu beside the search field. **Delete all chats** removes every chat you have in the project, and **Delete chats found** removes the chats that match your search or **Status** filter.
* **[Automatic Ask Oqo setup](https://docs.oqoqo.ai/operate/ask-oqo.md#open-ask-oqo):** Opening Ask Oqo in the dashboard for the first time connects it to your account automatically, with no separate setup step.
* **[Model picker in Ask Oqo](https://docs.oqoqo.ai/operate/ask-oqo.md#prerequisites):** Ask Oqo uses the same model picker as the other Oqoqo AI features, with models grouped by provider and the effort beside each one. While models load or need setup, your chats and draft stay open, and only sending waits.
* **[Compatible providers on agent pages](https://docs.oqoqo.ai/setup/agents.md#turn-on-an-agent):** Once an agent has a provider, its page lists other compatible providers under **Also works with**, and its row on **Manage → Agents** links there with **Connect more**. The CLI and the plugin list each agent's compatible providers too.
* **[GitHub source details on skills and files](https://docs.oqoqo.ai/experiments/tools.md#see-where-an-import-came-from):** The info icon on a skill or folder imported from GitHub shows the **Repository**, **Branch or tag**, **Path**, and **Commit** of the saved snapshot. The CLI and the plugin return the same details.
* **[Collapsible sidebars and a full-page Ask Oqo](https://docs.oqoqo.ai/reference/dashboard.md):** Double-click the edge of the sidebar or a side panel, or drag the edge past its narrowest width, to collapse it. In the Ask Oqo panel, select **Open full page** to continue the same chat on the full page, with your unsent message and attached files.

**November 17, 2026**

**New**

* **[API keys for the Oqoqo MCP server](https://docs.oqoqo.ai/operate/plugin.md#sign-in-with-an-api-key):** Connect an AI assistant using a personal API key via `OQOQO_API_KEY` in CI and other headless environments where browser sign-in isn't available.

**Improved**

* **[Faster model list loading](https://docs.oqoqo.ai/setup/model-providers.md#turn-off-or-refresh-models):** Models appear immediately after you connect or refresh a provider while Oqoqo checks which ones your credentials can use.
* **[Git history in task files](https://docs.oqoqo.ai/experiments/machines-and-files.md#git-history):** Uploaded task folders that include a `.git` directory preserve their commit history and branches, so agents can inspect past commits.
* **[Filtered model lists](https://docs.oqoqo.ai/setup/model-providers.md#which-models-oqoqo-keeps):** Embedding, image, and speech models are excluded from model lists, and each provider lists models your credentials can't use along with the reason.

**Fixed**

* Ask Oqo requests in Zed, JetBrains IDEs, and other ACP editors no longer time out after 30 seconds.

**November 16, 2026**

**New**

* **[Agent mode for Insights (private beta)](https://docs.oqoqo.ai/evaluation/agent-mode.md):** Oqoqo's insights agent reviews full runs without condensing, checks each friction against tool outputs, and shows what failed and passed runs did differently on **Comparisons**.

**Fixed**

* Loading placeholders are visible in dark mode again.

**November 15, 2026**

**Improved**

* **Updated brand palette:** The dashboard, emails, and public experiment links use the ink and purple palette from oqoqo.ai.

**November 14, 2026**

**New**

* **[Comparisons](https://docs.oqoqo.ai/evaluation/comparisons.md):** See which differences between treatments and agents hold up across tasks, and run likely differences again with the fewest runs needed to settle them.
* **[Brief](https://docs.oqoqo.ai/evaluation/comparisons.md#brief):** When Insights are on, each difference on **Comparisons** includes a one-sentence summary and the frictions seen on the side that did worse.
* **[Your own OAuth app for MCP servers](https://docs.oqoqo.ai/experiments/mcp-server-sign-in.md#connect-with-your-own-oauth-app):** Connect MCP servers that require a registered OAuth app, such as HubSpot, using your own Client ID and Client secret.

**Improved**

* **[Machine setup failure details](https://docs.oqoqo.ai/reference/troubleshooting.md#a-run-could-not-start-because-a-setup-command-failed):** Runs that stop during machine setup report the failed command along with its exit code and output.
* **[Variable conflict checks at launch](https://docs.oqoqo.ai/experiments/machines-and-files.md#when-two-places-set-the-same-variable):** Conflicting environment variable names across machines, tools, and providers are caught before any run starts.

**Fixed**

* Runs where the agent never started no longer report **Agent working** time.

**November 13, 2026**

**New**

* **[Turn off individual MCP tools](https://docs.oqoqo.ai/experiments/treatments.md#turn-off-mcp-tools):** Disable specific tools on an MCP server within a treatment to measure each tool's individual impact.

**Improved**

* **[Combined launch blocker checks](https://docs.oqoqo.ai/experiments/launch.md#launch):** Launch preflight reports every blocker at once instead of stopping at the first one.
* **[Address-bound MCP secrets](https://docs.oqoqo.ai/experiments/tools.md#saved-header-values-and-the-server-address):** Saved MCP header values and OAuth client secrets are sent only to server addresses approved in the dashboard.

**Fixed**

* Opening a file in the file viewer no longer returns an error.
* Imported skills use the `name` field from `SKILL.md` instead of their folder name.

**November 12, 2026**

**New**

* **[Agent mode for Evals (private beta)](https://docs.oqoqo.ai/evaluation/agent-mode.md):** Oqoqo's evals agent inspects full runs and changed files, verifies claims, and cites specific steps and diffs in each criterion verdict.
* **[Experiments, Runs, and Frictions tabs](https://docs.oqoqo.ai/evaluation/frictions.md#friction-lists):** Open any agent, task, tool, treatment, machine, file, or schedule to view every experiment, run, and friction tied to it.

**Improved**

* **[CI exit codes for watch and preflight](https://docs.oqoqo.ai/cli/experiments.md):** Use `oqoqo experiments watch --fail-on` and `oqoqo experiments preflight --require-ready` to fail CI jobs on specific terminal states or unready plans.

**Fixed**

* Runs stopped by internal Oqoqo errors are no longer listed as product frictions.

**November 11, 2026**

**New**

* **[Ask Oqo in Slack](https://docs.oqoqo.ai/operate/ask-oqo-in-slack.md):** Mention `@Oqoqo` in a Slack channel or send a direct message to launch experiments, check runs, and read results with your team.

**Improved**

* **[Follow-ups while Ask Oqo works](https://docs.oqoqo.ai/operate/ask-oqo.md#while-ask-oqo-works):** Send follow-up instructions while Ask Oqo is working instead of waiting for its current reply to finish.

**November 10, 2026**

**Improved**

* **[Friction fix targets and suggestions](https://docs.oqoqo.ai/evaluation/frictions.md#where-to-fix):** Each friction identifies where to fix it (**Product**, **Custom agent**, **Setup**, or **Agent behavior**) and includes a suggested fix.

**November 9, 2026**

**New**

* **[Schedules](https://docs.oqoqo.ai/experiments/schedules.md):** Launch finished experiments automatically on an hourly, daily, weekly, monthly, or custom cron schedule to track agent performance over time.
* **[Launch again](https://docs.oqoqo.ai/experiments/launch.md#launch-again):** Rerun a finished experiment in one step with the same settings, or select **Edit first** to adjust its configuration.

**Improved**

* **[Native default effort](https://docs.oqoqo.ai/setup/agents.md#choose-a-model-and-effort):** At **Default** effort, Oqoqo leaves the effort level unset so each agent uses its own vendor default.

**November 8, 2026**

**New**

* **[Tags](https://docs.oqoqo.ai/experiments/tags.md):** Label tasks, tools, treatments, and experiments with tags, and filter lists or experiment selections by tag.

**November 7, 2026**

**New**

* **[All frictions](https://docs.oqoqo.ai/evaluation/frictions.md):** Browse a project-wide list of every friction agents encountered, ranked by unrecovered runs, with dedicated detail pages for each friction.

**Improved**

* **[Custom model context windows](https://docs.oqoqo.ai/setup/model-providers.md#set-the-context-window-for-a-model):** Set a model's context window manually when a provider or gateway omits it, so Evals and Insights size their reads accurately.

**November 6, 2026**

**Improved**

* **[Live status in Ask Oqo](https://docs.oqoqo.ai/operate/ask-oqo.md#while-ask-oqo-works):** Follow what Ask Oqo is doing in real time with a live status line and collapsible tool-call cards.

**November 5, 2026**

**Fixed**

* MCP servers that use OAuth app credentials, such as HubSpot, stay connected after sign-in.

**November 4, 2026**

**Fixed**

* Deleting an official MCP server, CLI, or SDK completes properly.

**November 3, 2026**

**Improved**

* **[Streaming status in the CLI and plugin](https://docs.oqoqo.ai/cli/experiments.md#oqoqo-experiments-status):** Status reads can wait for the next state change, and result summaries indicate when Evals or Insights are still pending.

**Fixed**

* Experiment progress counts match across the dashboard, CLI, and plugin.

**November 2, 2026**

**New**

* **[GitHub repository browser](https://docs.oqoqo.ai/experiments/tools.md#import-from-github):** Browse and select a public GitHub repository, branch, and folder directly when importing skills or files instead of pasting a URL.

**Improved**

* **[Provider quota error handling](https://docs.oqoqo.ai/setup/model-providers.md):** When a model provider account runs out of usage or credits, runs stop without retrying and name the affected provider account.

**Fixed**

* Tasks mentioning rate limits are no longer flagged as provider rate-limit errors.

> [accordion] October 2026
>
> **October 31, 2026**
>
> **Fixed**
>
> * Skill and folder pages imported from GitHub open without errors.
> * OpenClaw runs on models that don't support an effort level.
>
> **October 30, 2026**
>
> **New**
>
> * **[GitHub repository reading in Ask Oqo](https://docs.oqoqo.ai/operate/ask-oqo.md#read-github):** Ask Oqo can read files, commits, releases, issues, and pull requests in public GitHub repositories while helping you write tasks or investigate frictions.
>
> **Improved**
>
> * **[Agent reasoning summaries in transcripts](https://docs.oqoqo.ai/evaluation/traces.md):** Run transcripts preserve reasoning summaries from Claude Code, Codex, GitHub Copilot, Hermes, and OpenClaw.
>
> **October 28, 2026**
>
> **Improved**
>
> * **Rerunning failed [Evals](https://docs.oqoqo.ai/evaluation/evals.md#run-evals-again) and [Insights](https://docs.oqoqo.ai/evaluation/insights.md#run-insights-again):** When Evals or Insights fail, completed runs are preserved so you can rerun them without repeating the experiment.
>
> **October 25, 2026**
>
> **Improved**
>
> * **[Ask Oqo speed, history, and approvals](https://docs.oqoqo.ai/operate/ask-oqo.md):** Chats start faster, keep per-project history, and ask for your approval before destructive actions or experiment launches.
>
> **October 24, 2026**
>
> **New**
>
> * **[Custom agents (private beta)](https://docs.oqoqo.ai/setup/custom-agents.md):** Connect any Agent Client Protocol (ACP) agent and evaluate it alongside Oqoqo's built-in agents.
> * **[Personal API keys](https://docs.oqoqo.ai/setup/profile.md#api-keys):** Create personal API keys under **Profile → API keys** to authenticate the CLI in CI and other non-interactive environments.
>
> **Improved**
>
> * **[Tools and Treatments terminology](https://docs.oqoqo.ai/experiments/tools.md):** Skills, MCP servers, CLIs, and SDKs are grouped under **Tools**, and combinations of tools are called **Treatments (Toolsets)**.
> * **[Run cost attribution](https://docs.oqoqo.ai/setup/custom-agents.md#where-a-cost-comes-from):** Runs show whether their cost was reported directly by the agent or estimated from token counts and list prices.
> * **Revoke Ask Oqo access:** Disconnect Ask Oqo from your organization from **Profile → Integrations**, the CLI, or the plugin at any time.
>
> **Fixed**
>
> * Transient network errors during agent installation retry automatically, and genuine install failures are reported clearly.
>
> **October 19, 2026**
>
> **New**
>
> * **[Ask Oqo in ACP editors](https://docs.oqoqo.ai/operate/ask-oqo-in-editors.md):** Chat with Ask Oqo directly from Zed, JetBrains IDEs, and other editors that support the Agent Client Protocol using `oqoqo acp`.
>
> **Improved**
>
> * **[Evidence coverage notes](https://docs.oqoqo.ai/evaluation/evals.md#what-evals-read):** When a run exceeds the model's context window, Evals and Insights show a note detailing which steps or files were shortened or omitted.
>
> **October 15, 2026**
>
> **New**
>
> * **[Run profile](https://docs.oqoqo.ai/evaluation/run-profile.md):** Every run includes a side rail summarizing treatment usage, tool call counts, notable patterns, and phase timing.
> * **[Run timeline](https://docs.oqoqo.ai/evaluation/traces.md#timeline):** Inspect when each tool was called across turns, and select any column to jump to that turn in the transcript.
>
> **Improved**
>
> * **[Actionable error messages](https://docs.oqoqo.ai/reference/troubleshooting.md):** Error messages across the dashboard explain what failed and link directly to the fix.
> * **[Background model list refreshes](https://docs.oqoqo.ai/setup/model-providers.md#turn-off-or-refresh-models):** Provider model lists refresh automatically in the background and notify you only when models are added, removed, or fail to update.
>
> **October 8, 2026**
>
> **Fixed**
>
> * Skill and MCP tool usage is counted accurately across all agents.
>
> **October 3, 2026**
>
> **Fixed**
>
> * Failed MCP server connections can be fixed and retried from the dashboard, CLI, and plugin.

> [accordion] August 2026
>
> **August 29, 2026**
>
> **New**
>
> * **[Notification preferences](https://docs.oqoqo.ai/setup/notifications.md#choose-your-notification-preferences):** Choose which experiment, billing, and model updates reach you by email and in the dashboard.
> * **[Advanced rubric mode](https://docs.oqoqo.ai/experiments/tasks.md#write-the-rubric):** Customize how individual criteria are scored and combined to pass a task when yes/no checks aren't enough.
>
> **Improved**
>
> * **[Per-tool timing and token totals](https://docs.oqoqo.ai/evaluation/run-profile.md#time):** Run details report exact duration for each tool call alongside total token usage.
>
> **August 22, 2026**
>
> **New**
>
> * **[Public experiment links](https://docs.oqoqo.ai/evaluation/share.md):** Publish read-only links to finished experiments so customers or teammates can inspect results without an Oqoqo account.
> * **[Qwen Code and additional providers](https://docs.oqoqo.ai/setup/model-providers.md#provider-list):** Run experiments with Qwen Code, and connect DeepSeek, Kimi, MiniMax, Alibaba Cloud Model Studio, and other providers with your own API keys.
>
> **Fixed**
>
> * Models from a Claude subscription load and run reliably.
>
> **August 18, 2026**
>
> **New**
>
> * **[Custom instructions for Evals and Insights](https://docs.oqoqo.ai/setup/oqoqo-ai.md#add-custom-instructions):** Add organization-wide or per-experiment guidance to tailor how Evals score runs and how Insights report frictions.
>
> **Improved**
>
> * **[Run output viewer and diff downloads](https://docs.oqoqo.ai/evaluation/traces.md#output):** Read each agent's final answer and changed files directly in the dashboard, or download the diff.
> * **Responsive dashboard layout:** Core dashboard views adapt cleanly to mobile and smaller screens.
>
> **August 15, 2026**
>
> **New**
>
> * **[Antigravity](https://docs.oqoqo.ai/setup/agents.md):** Evaluate Google's Antigravity agent on your tasks using a Google Gemini API key.
> * **[Cloudflare AI Gateway and GMI Cloud](https://docs.oqoqo.ai/setup/model-providers.md#provider-list):** Connect Cloudflare AI Gateway and GMI Cloud under **Organization settings → Model providers** to route multiple models through one key.
>
> **Improved**
>
> * **Refined dark mode palette:** Dark mode uses deeper neutral backgrounds for higher contrast and reduced glare.
> * **[Onboarding prompt on Home](https://docs.oqoqo.ai/setup/organization.md):** The Home page prompts you to connect a model provider first when none is configured.
>
> **August 13, 2026**
>
> **New**
>
> * **[Grok Build and xAI provider](https://docs.oqoqo.ai/setup/agents.md):** Evaluate xAI's Grok Build agent on your tasks using the new xAI model provider.
>
> **August 12, 2026**
>
> **Improved**
>
> * **[Matrix axis transpose](https://docs.oqoqo.ai/evaluation/results.md#matrix):** Swap rows and columns in the Matrix to read results either by task or by agent and treatment.
>
> **Fixed**
>
> * Organization names with accented or non-Latin characters save properly.
> * GitHub imports no longer fail for shared repositories.
>
> **August 8, 2026**
>
> **Oqoqo is open to everyone.**
>
> Write the tasks that matter to your product, run AI agents on them with and without your tools, and see what helps.
>
> * **[Experiments](https://docs.oqoqo.ai/quickstart.md):** Run AI agents such as Codex, Claude Code, and Cursor on your tasks in managed cloud machines.
> * **Evals and Insights:** Score every run against your rubric with [Evals](https://docs.oqoqo.ai/evaluation/evals.md), and surface product frictions with [Insights](https://docs.oqoqo.ai/evaluation/insights.md).
> * **Four ways to work:** Use the dashboard, the [Oqoqo Plugin](https://docs.oqoqo.ai/operate/plugin.md) in your AI assistant, the [CLI](https://docs.oqoqo.ai/operate/cli.md), or [Ask Oqo](https://docs.oqoqo.ai/operate/ask-oqo.md).
> * **[Monthly free plan](https://docs.oqoqo.ai/setup/billing.md):** Every organization starts on a free plan that resets its runs each month, with paid plans and non-expiring top-ups on [Pricing](https://oqoqo.ai/pricing).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.