Automation & Ops multi-agentself-hostedchat-tuiwhatsapptelegrammcp-memorycron-scheduler

Brigade — Enterprise-grade personal intelligence

A self-hosted crew of AI agents with long-term memory and a real org chart, reachable from terminal, WhatsApp, or Telegram.

FollowAgents review · FARS-2.1
Not recommended
59/ 100 5-point scale 3.0 / 5
1 2 3 4 5 6
1Trust17 / 29 · 2.9/5

Evidence shows: README and SECURITY.md describe permission controls (e.g., bash tool approval, ownership gates, keys stored at 0600, optional gateway tokens, AES-256-GCM encryption), but no code-level verification. Deductions: static review cannot confirm implementation details, and some claims (e.g., 'no telemetry') lack evidence.

2Reliability8 / 14 · 2.9/5

Evidence shows: README describes a coherent architecture (gateway, clients, memory, tools), and SECURITY.md provides a trust model. Deductions: no concrete examples of failure messages, and dependency availability is only indirectly indicated via version ranges in package.json.

3Adaptability12 / 18 · 3.3/5

Evidence shows: README describes multiple usage scenarios (terminal, messaging channels, MCP), capability boundaries (e.g., depth limits, approval gates), and trigger precision (e.g., cron expressions). Deductions: limited details on environment fit (e.g., Windows, Raspberry Pi).

4Convention10 / 18 · 2.8/5

Evidence shows: README is well-structured, includes installation notes, examples, known limitations (e.g., out-of-scope in SECURITY.md), and MIT license. Deductions: no changelog provided, naming stability not explicit, maintenance responsibility only indirectly via SECURITY.md and CI workflows.

5Effectiveness9 / 13 · 3.5/5

Evidence shows: README describes output usability (e.g., document generation, video rendering), marginal value (e.g., cross-model continuity, autonomous loops), and cost-benefit (e.g., bring your own model, no SaaS). Deductions: lack of actual performance or cost data.

6Verifiability3 / 8 · 1.9/5

Evidence shows: README and SECURITY.md provide some claims, but lack traceable test results or independent verification. Deductions: claims are mostly assertions without test reports or benchmark data.

Evidence confidence: Low Reviewed Sep 07, 2026 Reviewed revision 18ddd51d5316
Before you use it
  • Static review cannot verify security claims; dynamic testing is recommended.
  • Publisher identity is unverified; proceed with caution.
  • Many dependencies; supply chain security should be monitored.
Review evidence [1][2][3][4][5][6]
See the full review method →

What does this agent do, and when should you use it?

Brigade is a self-hosted crew of AI agents designed as an 'ecosystem, not an app' — an installable CLI (npm package `@spinabot/brigade`) whose gateway process serves multiple isolated agents with shared long-term memory (Tideline engine), a skill system, a cron scheduler, sub-agent fan-out, and an org hierarchy. It offers a terminal TUI (`brigade tui`) and a connect client over WebSocket, plus channel adapters for WhatsApp, Telegram, Slack, Discord, iMessage, and BlueBubbles. All data lives under `~/.brigade/` by default, with an optional fully self-hosted Convex backend. The agent supports many model providers (Anthropic, OpenAI, Gemini, Ollama, etc.) and can sign in with a subscription (Claude, ChatGPT, Copilot) or reuse a CLI login, eliminating API key management. Brigade claims no telemetry, and provides token-based gateway authentication, AES-256-GCM at-rest encryption, and permission gating on unsafe tools.

The core is a gateway process (brigade gateway run) that runs the agent loop, handling tool calls, approvals, and session-scoped context, bound to localhost:7777 by default. The TUI (brigade or brigade tui) and brigade connect attach to this gateway over WebSocket. Agents are provisioned through CLI utilities: brigade onboard walks through storage mode, provider, and model; brigade agents add creates isolated agents with separate workspaces and sessions; brigade org init defines an org chart and cross-agent communication policy. Built-in tools include file editing (read/write/edit/bash/grep), memory tools (recall_memory, etc.), sub-agent spawning, web_search, analyze_media for reading PDFs/Office docs/image/audio/video, render_video for HTML-to-MP4 motion graphics, and 1,000+ app connectors via Composio. It also exposes brigade mcp to serve long-term memory to MCP clients, brigade cron for scheduling, and brigade expose for secure public tunneling.

  1. Terminal-first developers who want coding assistance on a local machine or Raspberry Pi with their own API key and a local Ollama model.
  2. Small teams that deploy a self-hosted gateway and want members to interact with isolated agents via WhatsApp or Telegram, with shared memory but separate credentials.
  3. Users who need a personal assistant with persistent memory across sessions — storing preferences and reminders — and access it from Slack, Discord, or iMessage.
  4. Ops users who schedule recurring agent turns or system events with brigade cron to get daily reports or reminders delivered to a channel.
  5. Developers who use MCP clients and want to expose a custom knowledge base through brigade mcp.
  6. Privacy-conscious individuals who want full data ownership with optional at-rest encryption and no cloud dependency.

What are this agent's strengths and limitations?

Pros
  • Self-hosted with zero telemetry; data defaults to ~/.brigade/ and optional Convex mode adds at-rest encryption.
  • Tideline memory engine provides provenance-gated writing, hybrid BM25+vector recall, and bi-temporal decay.
  • Keyless sign-in via Claude/ChatGPT/Copilot subscriptions or reuse of Claude Code/Codex CLI logins.
  • Multi-agent isolation, org chart policy, and a single binary that also acts as an MCP server.
Limitations
  • Requires Node and CLI comfort; no managed dashboard or hosted option.
  • You must bring your own model subscription/API key unless you run local Ollama (limited).
  • Feature-dense documentation might overwhelm beginners; learning curve is steep.
  • Privacy and no-telemetry claims are not backed by an independent third-party audit.

How do you install or deploy this agent?

Requires a shell on macOS/Linux or PowerShell on Windows, plus Node 22.12+ or the official installer. On macOS/Linux, run curl -fsSL https://brigade.spinabot.com/install.sh | sh (installs Node automatically if needed). On Windows PowerShell, run irm https://brigade.spinabot.com/install.ps1 | iex. If you already have Node, you can npm i -g @spinabot/brigade.

How do you use this agent?

Run brigade onboard and follow the 5-step wizard to choose storage mode (filesystem or self-hosted Convex), provider, and model. Then run brigade tui (or just brigade) to start the chat UI, which auto-starts the gateway. For always-on operation, run brigade gateway install to install a system service. Use brigade agents add <name> to add extra agents. Common commands: brigade agent -m "summarize ~/today.md", brigade cron add --cron "0 9 * * *" --message "good morning", and brigade channels link --channel whatsapp.

FAQ

How is Brigade different from a single cloud AI assistant?
Brigade is a self-hosted installation that lets you run multiple isolated agents, each with their own memory and credentials, coordinated through a real org chart, with no hosted API in the middle. It is not a single assistant; it scales to a crew that can be used for team or operational workflows.
Do I need to bring my own API key or subscription?
Yes, by default. You either provide a provider key (Anthropic, OpenAI, Gemini, etc.), use local Ollama, or sign in with your Claude/ChatGPT/Copilot subscription via keyless OAuth. Brigade does not host models itself.
How private is my data?
All state is stored locally in ~/.brigade/ by default, keys are chmod 0600, the gateway binds to 127.0.0.1 by default, optional token authentication is available, and optional AES-256-GCM encryption at rest in self-hosted Convex mode. There is no telemetry. However, you may want to audit it yourself; no independent verification exists yet.
Can it run on low-powered hardware?
Yes, the docs say the same code runs on a Raspberry Pi or a server. It requires Node.js 22+ and disk space. No minimum RAM or CPU is specified, but the TUI is lightweight and the heavy lifting is done by the connected model and tools.

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Related agents