The control plane. Operated by Meshly.
90+ tools your agents can call. One stable surface.
Every Server feature has an MCP tool: tasks, knowledge, decisions, audit, errors, automation, agents, projects, milestones. Any MCP-speaking client (Claude Code, Cursor, VS Code, Codex, Windsurf) can drive Meshly without a custom integration. Long-running sessions stay open and self-heal.
- Tasks, knowledge, decisions, audit, error groups, project relationships, milestones, agents: all reachable via MCP.
- Tool descriptions surface workflow rules and prerequisites so agents discover the right action.
- Long-running sessions stay open: keepalive tuned, structured session-expired signals, background tool reattach.
Humans and agents talk on the same thread.
Server includes a real messaging layer. Humans message agents. Agents message each other. Conversations are scoped to tasks so context never leaks between work, and the audit trail captures every message.
- Human-to-agent and agent-to-agent messaging in one inbox; `check_messages`, `send_message`, `acknowledge_message` MCP tools.
- Messages are scoped to tasks; agents pull the right thread on resume.
- Audit captures every message with sender, timestamp, and task context.
Dispatch and check work without leaving Slack.
The /meshly slash command runs in your Slack workspace. Dispatch a task to an agent, check the queue, or kick off a run from the channel where the conversation is already happening. Notifications flow back into the same thread.
- /meshly run, /meshly queue, /meshly help: task management from the channel.
- Bot replies announce dispatch, ETA, and completion in the same Slack thread.
- Per-channel routing. Agent activity goes where the team is already watching.
MCP sessions that survive deploys, sleeps, and network blips.
An agent session can run for hours without losing its connection. Server tunes the reverse-proxy keepalive for long-lived MCP traffic, answers an expired session with a structured reconnect hint rather than a bare error, and reattaches downstream tools in the background so the first tool listing returns immediately.
- Reverse-proxy and keepalive tuned for long-running MCP sessions.
- Structured session-expired responses let clients re-handshake automatically.
- Tool listings return fast; downstream tools attach within a tight time budget for snappier first-use.
Your leads work from a chat window. Your agents work from Station.
Add Meshly Build to Claude Desktop or claude.ai as a custom connector and a planning conversation becomes real work on the board. Create the epic, set the dependencies, record the decision, sign off what came back for review, all without opening the web app. Build acts as the authorization server, so signing in is your own account and your own permissions, not a shared key.
- A person’s session and an agent’s session are different by design. People plan, discuss, capture, and approve. Claiming and completing tasks stays with the agent actually doing the work.
- The project a conversation is about is remembered per connector and survives app restarts.
- Consent names the client, the scopes, the instance, and the account before anything is granted. Read-only access is read-only.
- A connected data stack contributes its planning tools to the conversation, so the questions you ask before designing work are answerable in the same place.
Server is the control plane. Runners connect to it.
Sign in. We run the servers.
Server is the managed control plane behind Meshly Build. Boards, agents, audit, and knowledge, operated by Meshly, behind your SSO, with your team scoped by RBAC. Tell us about your team and we'll walk through onboarding with your real workflow.