Why BMO
Use this page when you are choosing a coding agent, not operating one yet. For install and first run, go to Quickstart. For day-to-day routing after adoption, see Decision guide.
What BMO optimizes for
Section titled “What BMO optimizes for”BMO is a terminal-native, local-first, BYOK coding agent. You run the binary on your machine; prompts and selected file or tool content go to your configured provider. Other destinations depend on the tools, execution backend, and telemetry you enable; see Security and data.
The product bet is inspectable agent work: shared session state,
policy controls on tool execution, durable run evidence, and one agent core across
bmo (TUI), bmo run, HTTP/MCP/ACP servers, and A2A — not separate products
per surface.
Compare the workflow you need
Section titled “Compare the workflow you need”Use the same repository task, approval settings and evidence requirements when comparing tools. Product defaults vary by version and configuration; this page describes BMO’s surfaces rather than ranking other products.
| Evaluation question | What to inspect in BMO |
|---|---|
| Can I work from my terminal and checkout? | TUI and CLI operate on the workspace you select. |
| How are mutations authorized? | Session mode, tool policy and grants determine whether a call is allowed, denied or prompts. |
| Can I inspect a past run? | Local sessions, run evidence and the debugger; inspect the exact records your workflow needs. |
| Which integration fits my client? | CLI, HTTP, MCP and ACP have distinct transport and authorization contracts. |
| Where do commands run? | The sandbox backend selects the host or configured execution environment; local is not isolation. |
| How do I reach a private MCP server? | Supply the network path; BMO does not provide a managed tunnel. |
| What will it cost? | Provider usage plus any configured infrastructure and tool-service charges. |
BMO is a fit for terminal workflows with inspectable local state and explicit
policy controls. It also integrates with editors: ACP (bmo acp) for
Zed-style agent panels and MCP for context tools — see
Zed and Protocols. A
managed or turnkey hosted IDE product, bundled model access, or managed
network connectivity may still fit better when you want that whole stack
hosted for you rather than a local binary plus BYOK.
The conceptual synthesis diagrams describe a broader destination. They are not comparison results or evidence that all depicted capabilities work together in a generally qualified deployment.
Use both when an IDE agent handles in-editor completion and BMO handles repo-wide refactors, headless CI, or ACP/MCP-exposed agent work — see Cursor and IDEs.
Evaluation questions (quick answers)
Section titled “Evaluation questions (quick answers)”| Question | Answer |
|---|---|
| Is it free? | The binary has no license fee. Redistribution follows the terms shipped with a published release. You pay configured provider, infrastructure, and tool-service charges. See Cost and operations. |
| Is it stable? | Pin release tags for stability-sensitive use. See Limitations and maturity map. |
| Can I audit runs? | Yes — sessions, tool calls, and debugger timeline persist locally. See Security and data. |
| Multi-agent on day one? | No — start solo; advance when needed. See Decision guide. |
Related
Section titled “Related”- Overview — fit table and book navigation
- Protocols — surface inventory
- Security posture — technical summary