@mtreeai/msapling-cli 2.3.6-beta.50 → 2.3.6-beta.52

package/README.md

@@ -61,13 +61,18 @@ Once the REPL is running, type `/help` to see all available slash commands. Comm
 | `/login <token>` | Paste an API token from Settings → API Keys |
 | `/logout` | Clear stored token |
 | `/chat <message>` | Send a message and stream a response |
-| `/swarm <prompt>` | Run parallel agent swarms |
-| `/models` | List available models |
+| `/chat new <title>` | Create a chat inside the fixed CLI project |
+| `/chat import <chat_id>` | Copy another owned chat into a new CLI chat without mutating the source |
+| `/model` | Open the loopback visual catalog (complete LAB catalog remotely; direct provider catalog standalone) |
+| `/model --terminal` | Use the provider-first terminal selector instead |
+| `/keys add <provider>` | Enter a BYOK key through hidden terminal input or the one-time local browser prompt |
+| `/keys use <provider> [base_url]` | Select OpenRouter or an OpenAI-compatible provider in standalone mode |
+| `/footer [compact|expanded|field on|off]` | Configure telemetry visibility |
 | `/cost` | Show usage and billing data |
 | `/config` | View and edit configuration |
 | `/help` | Show all slash-command help |
 
-> **Note:** `chat`, `login`, `models`, `cost`, `swarm`, and `config` are REPL slash-commands (prefixed with `/`), not top-level argv subcommands. Run `msapling --help` to see the argv interface.
+> **Note:** `chat`, `login`, `model`, `footer`, `cost`, and `config` are REPL slash-commands (prefixed with `/`), not top-level argv subcommands. The CLI is single-chat: fanout/swarm commands are hidden and disabled. Run `msapling --help` to see the argv interface.
 
 See `packages/cli/src/commands/index.ts` for the canonical slash-command list.
 
@@ -75,7 +80,7 @@ See `packages/cli/src/commands/index.ts` for the canonical slash-command list.
 
 
 
-The CLI is not meant to be a large standalone agent runtime. MSapling's backend is the source of truth for auth, projects, chats, model routing, usage, billing, MDrive, locks, and audit data. The website, mobile app, and CLI should all show and modify the same backend state.
+The installed CLI has explicit remote and standalone profiles. In remote mode, MSapling's backend is the source of truth for auth, projects, chats, model routing, usage, billing, MDrive, locks, and audit data. In `--local` mode, model calls, provider keys, transcripts, tools, and project settings remain local and LAB is not contacted for turns.
 
 ## Product Direction
 
@@ -83,7 +88,7 @@ MSapling started as a multi-chat browser workspace: multiple AI chats, many mode
 
 The mobile app extends that shared workspace to one-chat-at-a-time usage and phone-first workflows such as sending a photo into a chat.
 
-The CLI brings the same backend workspace to the terminal. It should be fast, low-resource, and useful without duplicating the backend.
+The CLI uses the same backend state through a fixed project named `CLI` and one active chat at a time. It should be fast, low-resource, and useful without duplicating the backend.
 
 ## Architecture
 
@@ -96,7 +101,7 @@ CLI         /
 The CLI should:
 
 - authenticate with the backend
-- read projects and chats from the backend
+  - read the fixed `CLI` project and active chat from the backend
 - stream chat responses from `/api/chat/message`
 - display backend usage and quota data
 - respect backend stream locks
@@ -105,7 +110,8 @@ The CLI should:
 
 The CLI should not:
 
-- invent separate chat state
+  - invent separate chat state or switch arbitrary web chats
+  - expose multi-chat/swarm fanout commands
 - compute billing locally
 - simulate MDrive state
 - bypass backend locks