> ## Documentation Index > Fetch the complete documentation index at: https://docs.braintrust.dev/llms.txt > Use this file to discover all available pages before exploring further. # Manage Loop threads and limits > Copy, export, share, rename, and delete threads, understand sandbox lifecycle, and learn how Loop limits concurrent active threads. Loop saves your threads so you can return to work in progress. Manage their history and access, and learn how sandboxes, concurrency limits, providers, and privacy settings affect your work. ## View thread history Long threads load recent history first. After Loop finishes working, scroll to the top and click **Load earlier messages** when available. ## Copy or export a thread A Markdown transcript of a thread includes your messages with the names of any attached context, Loop's responses, and each tool call with its arguments, output, and any error. Use it to share an investigation outside Braintrust or hand it to a coding agent. Attachments appear only as a count in the transcript and are not embedded. The transcript covers only loaded messages. To include earlier history, click [**Load earlier messages**](#view-thread-history) until it no longer appears. * **Copy to the clipboard**: Click **Copy as Markdown** in the thread header. In narrow layouts and in Loop chats opened from other pages, the button shows only its icon. * **Download a file**: On the [** Loop**](https://www.braintrust.dev/app/~/loop) page, select in the thread header, then **Export conversation** to download the transcript as a `.md` file. Only the thread owner can export a thread. ## Share a thread Threads are private by default. Share a thread you own so project members can review the entire conversation or continue the investigation. * Loop accesses Braintrust using each message sender's permissions. * Only the sender can approve or cancel work started by their message. * Only the owner can edit messages, and only their own. 1. Go to [** Loop**](https://www.braintrust.dev/app/~/loop) and select your thread. 2. Click **Share** in the thread header. 3. Choose an access level: * **Members can view** lets members with project read access view the conversation. * **Members can chat** also lets members who can create Loop threads send messages. 4. Click **Save** to apply the access level and copy the thread link. To copy the link again, reopen **Share** and click **Copy Link**. Only the owner can change access. Reopen **Share**, choose an access level or **Private to you**, and click **Save**. > [!WARNING] > > Before making a thread private or removing chat access, have participants finish or cancel their work. Changing access leaves their queued or running turns in progress but prevents them from approving actions or canceling those turns. ## Rename a thread Only the thread owner can rename a thread. 1. On the [** Loop**](https://www.braintrust.dev/app/~/loop) page, select in the thread header or beside a thread in the sidebar, then **Rename thread**. 2. Enter a name of up to 256 characters and click **Save**. ## Delete a thread Only the owner can delete a thread, and only while Loop isn't working in it. A sleeping sandbox doesn't prevent deletion. Automation run threads are read-only and cannot be deleted this way. See [Review past runs](https://docs.braintrust.dev/loop/automations.md#review-past-runs). > [!WARNING] > > Deleting a thread is permanent and cannot be undone. On the [** Loop**](https://www.braintrust.dev/app/~/loop) page, select in the thread header or beside a thread in the sidebar, then **Delete thread** and confirm. ## Sandboxes Loop uses an isolated, per-user sandbox to inspect trace files and run commands against your data. * **Start.** A sandbox starts with your first message in a thread. Opening the [** Loop**](https://www.braintrust.dev/app/~/loop) page or revisiting a saved thread doesn't start one. * **Sleep.** After inactivity, the sandbox sleeps. Your next message resumes it automatically with the thread history intact. * **Expiry.** A sandbox lasts up to 8 hours, including time asleep. After it expires, your next message starts a fresh sandbox without affecting the thread or its history. > [!NOTE] > > Self-hosted deployments provide their own sandboxes, with behavior depending on provisioned capacity. On AWS, enable the [Loop runtime](https://docs.braintrust.dev/admin/self-hosting/configure/loop-runtime.md) to run Loop in your data plane. Until then, Loop runs in the browser. ## Thread concurrency Braintrust-hosted Loop limits active threads separately for each member, service token, and automation within an organization. A thread occupies a slot while Loop works and for a short idle period afterward. Further messages in that thread don't consume extra slots. | Organization plan | Concurrent active threads | | - | - | | Starter without on-demand usage | 5 | | Starter with on-demand usage, Pro, and Enterprise | 50 | At the limit, starting work in another thread returns HTTP `429`. Wait for a slot to become available, then retry. Existing threads and histories are unaffected. ## Models and providers For any of [Loop's models](https://docs.braintrust.dev/loop.md#choose-a-provider-and-model), choose how inference is served and billed with the model picker's **Provider** setting. Loop remembers the selected provider for each thread. * **Braintrust.** Use Braintrust's AI providers with no setup. Usage draws from your organization's monthly [model credits](https://docs.braintrust.dev/plans-and-limits.md#model-credits), shared with Topics. On Starter, you need an organization owner with a work email address or a payment method on file. See [Requirements](https://docs.braintrust.dev/admin/ai-providers.md#requirements). * **Your configured provider.** Use an OpenAI-compatible [AI provider](https://docs.braintrust.dev/admin/ai-providers.md) configured for your organization or the thread's project, with inference billed to that provider. New threads start with **Auto** on, which follows your organization's [default provider](https://docs.braintrust.dev/admin/organizations.md#set-a-default-provider-for-loop) and its recommended model and reasoning effort. [Automations](https://docs.braintrust.dev/loop/automations.md) with **Auto** on follow the default on every run. If no default is saved, the default is **Braintrust** when built-in models are available. Picking another provider in a thread turns off **Auto** for that thread only, and the next new thread starts on the default again. If there's no default, or the saved default is unavailable, **Auto** is unavailable and new threads use your most recent choice. > [accordion] Configure your own provider > > Loop supports the following provider types and model configurations. You can add a provider from the model picker or manage access through your organization's settings. > > * **Compatibility.** Supported types are OpenAI, Azure, Bedrock, Databricks, OpenRouter, and Azure AI Gateway. Your provider must serve the selected Loop model. You need a configured provider once model credits run out. > > * **Model IDs.** Loop sends the model ID configured on your provider as is, including Bedrock cross-region IDs such as `us.openai.gpt-6-luna` and `global.openai.gpt-6-luna`. The ID must still serve one of Loop's models. > > * **Databricks.** Databricks uses its own IDs for Loop's models: `databricks-` plus the model name with dots changed to hyphens, such as `databricks-gpt-6-sol` or `databricks-gpt-6-1-sol`. On a Databricks provider, Loop doesn't recognize IDs such as `gpt-6-sol` or `openai/gpt-6-sol`. > > * **DeepSeek endpoints.** DeepSeek's API serves DeepSeek models, not Loop's GPT models, so Loop doesn't offer a provider whose **API base URL** host is `api.deepseek.com`, even if its type is supported. Gateways and proxies on other hostnames aren't excluded, but they must still serve the selected Loop model. > > * **Add a provider.** Click **Add provider** in the model picker, then **New** under **Custom providers**. The form offers compatible types. Loop also recognizes supported models with gateway provider prefixes. Custom endpoints and prefixes do not expand Loop's supported models. > > * **Restrict provider access.** To require your organization's providers, turn off **Allow built-in models** on the [AI providers](https://docs.braintrust.dev/admin/ai-providers.md#manage-built-in-models) page. This removes **Braintrust** from Loop's provider options. > > > [!NOTE] > > > > On self-hosted deployments, Databricks support and unchanged provider model IDs require data plane v2.19.0 or later. Earlier versions don't offer Databricks for Loop and don't use provider-specific IDs, such as Bedrock cross-region IDs. ## Privacy Threads are private by default. You can [share a thread with project members](#share-a-thread) to grant access to its conversation history and optionally let them send messages. Braintrust logs frontend Loop interactions for product improvement and support by default. You can [turn off logging](https://docs.braintrust.dev/admin/organizations.md#control-loop-logging) without affecting Loop. Self-hosted organizations have it off by default. Model data handling depends on the feature: * **Loop.** Your configured provider uses the key configured for that provider. Your agreement with that provider, including any zero data retention agreement, governs prompts and responses. See [Data handling](https://docs.braintrust.dev/admin/ai-providers.md#data-handling). * **Automations and Patterns.** [Loop automations](https://docs.braintrust.dev/loop/automations.md) and [Patterns](https://docs.braintrust.dev/observe/patterns.md) use the model and provider saved in each automation's configuration. * **Topics.** Facet summarization, embeddings, and cluster naming in [Topics](https://docs.braintrust.dev/observe/topics.md) always use Braintrust-served models with Zero Data Retention, regardless of your Loop model. ## Next steps * [What Loop can do](https://docs.braintrust.dev/loop/capabilities.md) for the full range of tasks. * [Run Loop on a schedule](https://docs.braintrust.dev/loop/automations.md) for work that shouldn't depend on an open chat. * [Plans and limits](https://docs.braintrust.dev/plans-and-limits.md) for the rest of the system limits. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.