bamboo-agent 2026.7.21

A fully self-contained AI agent backend framework with built-in web services, multi-LLM provider support, and comprehensive tool execution
Documentation
# Docs index

New here? Start at [`guides/GETTING_STARTED.md`](guides/GETTING_STARTED.md).
The top-level [`README.md`](../README.md) has the project overview, capability
summary, and CLI command table.

## User guides

Task-oriented, kept current with the shipped codebase.

| Doc | What it covers |
|---|---|
| [`guides/GETTING_STARTED.md`]guides/GETTING_STARTED.md | Install, configure a provider, run your first turn three ways (CLI/HTTP/SDK). |
| [`config-reference.md`]config-reference.md | Every `config.json` key, the sibling config files (`connect.json`, `schedules.json`, `model_limits.json`), every `BAMBOO_*` env var, the secret-masking contract, encryption at rest, and corrupt-config recovery. |
| [`guides/CONNECT.md`]guides/CONNECT.md | Drive Bamboo from Telegram or Feishu/Lark. |
| [`guides/PLUGINS.md`]guides/PLUGINS.md | Install/update/remove plugins, and the three-layer URL trust model. |
| [`guides/DEPLOY.md`]guides/DEPLOY.md | Run `bamboo serve` long-lived: bare binary, systemd, Docker, reverse proxy, backups. |
| [`guides/API.md`]guides/API.md | HTTP/SSE API reference. |
| [`guides/MIGRATION_GUIDE.md`]guides/MIGRATION_GUIDE.md | Upgrading across breaking changes. |
| [`architecture-crates.md`]architecture-crates.md | The Cargo workspace's crate layers and dependency direction — useful when extending Bamboo itself, not just using it. |
| [`claude-code-executor.md`]claude-code-executor.md | Protocol reference for driving the `claude` CLI as a sub-agent executor (`subagents.executor = "claude_code"` in the [config reference]config-reference.md#sub-agents--the-claude_code-executor). |
| [`security/KNOWN_VULNERABILITIES.md`]security/KNOWN_VULNERABILITIES.md | Tracked `cargo audit` advisories and their resolution status. |
| [`../examples/`]../examples/ | Compiling, runnable code — quickstart, streaming real event types, a custom tool, resuming a session, the `ExecuteRequest` escape hatch, connecting an MCP server. Built in CI (`cargo build --examples`), so they can't silently drift from the real API. |

## Design notes (historical)

[`design/`](design/) holds implementation plans and architecture RFCs written
*before* or *during* the work they describe — most of what they propose has
since shipped. They're kept for provenance (the "why" behind a design, ratified
decisions, alternatives considered) but are **not maintained** against the
current codebase and may describe an intermediate or superseded state. If a
design doc and the code disagree, the code is correct — read the doc for
history, not as a spec.

<details>
<summary>Full list</summary>

- [`design/api-v2-transport.md`]design/api-v2-transport.md — unified WS transport + in-process TLS design (API v2 transport epic).
- [`design/architecture-overview.md`]design/architecture-overview.md — broker-mediated remote sub-agents, hub-and-spoke topology.
- [`design/ask-agent-design.md`]design/ask-agent-design.md — durable mailbox request/reply to a sub-agent (`ask_agent`).
- [`design/edit-fuzzy-matching-design.md`]design/edit-fuzzy-matching-design.md — fuzzy matching design for the `Edit` tool.
- [`design/ergonomic-sdk-plan.md`]design/ergonomic-sdk-plan.md — the plan behind the current `bamboo-sdk` facade (`Agent`/`AgentBuilder`).
- [`design/feishu-adapter-plan.md`]design/feishu-adapter-plan.md — Feishu/Lark `bamboo-connect` adapter plan (epic #447 phase 3).
- [`design/multi-provider-routing-refactor-plan.md`]design/multi-provider-routing-refactor-plan.md — multi-provider/multi-model routing (session-level provider+model).
- [`design/personal-assistant-ledger.md`]design/personal-assistant-ledger.md — the ledger memory subsystem design.
- [`design/phase1-provider-registry-plan.md`]design/phase1-provider-registry-plan.md / [`phase1-provider-registry-patches.md`]design/phase1-provider-registry-patches.md`ProviderRegistry` multi-instance provider runtime, plan + patch breakdown.
- [`design/prompt-cache-architecture-plan.md`]design/prompt-cache-architecture-plan.md — prompt/message/cache reorganization for stable provider prompt-cache hits.
- [`design/provider-model-first-class-plan.md`]design/provider-model-first-class-plan.md — making provider+model a first-class selection unit.
- [`design/remote-actor-plan.md`]design/remote-actor-plan.md — remote sub-agent actor seams (P0/P1/P2).
- [`design/remote-mailbox-broker-design.md`]design/remote-mailbox-broker-design.md — the standalone `bamboo broker` network mailbox design.
- [`design/subagent-actor-runtime-design.md`]design/subagent-actor-runtime-design.md — the virtual-actor model sub-agents run on today.
- [`design/subagent-store-mailbox-interface.md`]design/subagent-store-mailbox-interface.md`bamboo-subagent`'s `store/`/`mailbox/` interface spec.
- [`design/reviews/actor-runtime-self-review-2026-06-12.md`]design/reviews/actor-runtime-self-review-2026-06-12.md — a self-review snapshot of the actor runtime.

</details>

Adding a new design doc? Put it in `design/` from the start if it's a plan/RFC
for work not yet shipped — promote genuinely user-facing outcomes into a guide
above once the feature lands, rather than leaving the plan doc as the only
documentation.