Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/archestra-ai/archestra/llms.txt

Use this file to discover all available pages before exploring further.

Archestra includes a built-in Chat interface for working with agents, MCP tools, files, browser actions, and model selection in one place. It is the fastest way to interact with a configured agent — select an agent, type a message, and the agent executes using its assigned tools and knowledge sources. Tool calls are surfaced inline as the agent works, so you can see exactly what it is doing at each step.

Selecting an Agent

Open Chat from the left sidebar. Use the agent selector at the top of the conversation to pick which agent handles your messages. Each agent brings its own system prompt, tools, and knowledge sources to the conversation.

Tool Use in Chat

When an agent calls a tool during a conversation, the call and its result appear inline in the chat thread. This gives you full visibility into what the agent looked up, ran, or modified while composing its response — including which knowledge sources were queried via query_knowledge_sources.

Suggested Prompts

Agents can be configured with suggested prompts — pre-written example messages that appear when you open a new conversation. These help users get started quickly with the most common tasks the agent is designed for. Suggested prompts are set on the agent’s edit page and are visible only to that agent’s users in chat.

File Attachments

You can attach files directly in the chat input. Attachments are scoped to their conversation by default. To reuse an attachment in future conversations or make it retrievable by agents, save it to Knowledge > Files:
  1. Upload or send a file in chat
  2. Click the save icon on the attachment
  3. Set the visibility (Owner, Teams, or Organization)
  4. Select the agents or MCP Gateways that can retrieve it
The knowledgeFile:create permission is required to save chat attachments to Knowledge Files.

Chat Commands

Type / in the message input to open available chat commands:
CommandDescription
/compactSummarizes older conversation history to reduce context usage and help prevent hitting the model’s context limit. The full chat history remains visible.

Context Compaction

Context compaction replaces older messages sent to the model with a structured handoff summary while keeping recent turns verbatim. In short conversations, Chat can summarize completed earlier work instead of waiting for four user turns.
  • The original chat history remains visible in the conversation
  • Compaction events appear in the conversation timeline
  • Uploaded text files and PDFs are included in the compaction transcript when extractable text is available
Compaction is handled by the built-in Context Compaction Subagent. Users with agent:admin permission can edit its instructions and model from the built-in agent settings. If no model is configured on the subagent, Chat uses the workspace default model, then the best available configured model.
Use /compact proactively in long sessions with large tool outputs to stay within model context limits without losing conversation continuity.

Chat Titles

Chat titles are generated by the built-in Chat Title Generation Subagent. Users with agent:admin permission can edit its instructions and model from the built-in agent settings. If no model is configured on the subagent, Chat uses the workspace default model, then the best available configured model.

Supported LLM Providers

Chat supports the LLM providers configured for your workspace. API keys are configured per-provider under Settings > LLM Providers. See Supported LLM Providers for the full list of supported providers and models.

Build docs developers (and LLMs) love