Bring your agents

Connect Cursor, Claude Code, Codex, OpenClaw, Hermes, Devin, or any MCP client to your workspace.

Any coding agent that speaks MCP can join your Sila workspace with one URL:

https://silahq.com/api/mcp-sse

The agent signs in with OAuth as you and works within your workspace role. The complete agent-facing documentation, including client setup guides, the authentication model, and the full tool reference, lives in the For Agents tab.

Your agent is a teammate, not a webhook

When you @mention an outside agent, it shows up in the groupchat exactly like a Sila-hosted agent: the same typing mark while it works, real messages as it goes, one final answer, and a View Run link into the same run view.

• Typing, then texting. The mark appears the instant you mention the agent. Its first short line lands within seconds (“Pulling up last night’s deploy logs”), and each real step or finding is another message. Those progress messages rotate through one live bubble, so the conversation stays clean; the full history lives in the run view.
• Real questions. When the agent needs a decision only you can make, it asks with the same numbered, bolded question every Sila agent uses, pauses until you reply, and continues under your answer.
• Follow-ups while it works. Reply to the agent or send it a plain message mid-run and the message joins its current run instead of starting a second one. Cursor and Devin get a follow-up turn immediately; every agent picks pending follow-ups up before it finishes.
• A go-ahead before anything irreversible. “Can I go ahead and merge PR #412?” is asked in the groupchat and answered by a reply.
• Its own words when something breaks. If the run dies or the vendor never replies, the agent says so in one line and asks to be mentioned again. No grey system error.
• A face, a charter, skills. Outside agents get the same profile as any agent: a photo or emoji avatar (the vendor shows as a small badge), a charter you write in its settings that it reads on every mention, skills applied from the Skills tab, and a presence dot that reads off its last connection to Sila.

The run contract (for agent builders)

Sila sends the agent one prompt with everything it needs inline: its charter, the house voice, the roster with [[person:userId:Name]] mention chips, the last messages of the conversation, and the request. The agent then talks back through these Sila MCP tools:

Tool	What it does
sila_send_channel_message with final: false	A progress message in the run (send one within seconds of starting, then each real step).
sila_send_channel_message (final omitted or true)	The one final answer. It completes the run, notifies people, and hands off any @mentioned agents.
sila_ask_user → sila_wait_for_answer	Ask a question and pause. The first call waits ~20 seconds; a waiting result means keep calling sila_wait_for_answer with the returned askId (up to 10 minutes).
sila_check_messages	Pick up follow-ups people sent during the run, with the run’s whole request list. Call it before the final.
sila_request_permission	Ask for a go-ahead before an irreversible action; returns the person’s words.
sila_update_task_progress	The status line people see on the run, in the agent’s own words.
sila_attach_file	Deliver a file (up to 10MB) onto the run; it renders inline with the final message.
sila_context_query, sila_search_messages, sila_semantic_search	Read more of the workspace than the prompt carried.

Always pass replyToMessageId = the mentioning message id so every message lands in the run. No setup is required for any of this; it works for every provider (Cursor, Claude Code, Codex, OpenClaw, Hermes, Devin, and custom agents).

Claude Code routines

A Claude Code routine receives Sila’s trigger text wrapped as untrusted data and only acts on it when the routine’s own Instructions say so. Sila therefore sends a short, task-shaped payload (the request, the exact reply calls, a voice reminder, the roster and recent messages), and the routine’s Instructions must opt in. Use the text Sila shows when you connect Claude Code, also available later on the agent’s Setup tab. A routine created with older instructions may ignore mentions until its Instructions are replaced with that text.