amness 0.1.0 → 0.1.3

package/docs/providers.md

@@ -64,11 +64,22 @@ The model list in `/model` (with prices per million tokens as shown in the picke
 
 **Limits and common problems.**
 
-- Amness saves a key you type in `~/.amness/config.json` with file mode 600 (only you can read it).
+- Amness saves a key you type in your system's secure store: the macOS Keychain, the Linux keyring (`secret-tool`) or Windows DPAPI. In Docker or Kubernetes it saves nothing: pass the key as an environment variable or a `*_FILE` secret. On a Linux server with no keyring it uses `~/.amness/config.json` with file mode 600, which is not encrypted. See [Safety](safety.md).
 - The environment variable wins over the saved file. `ANTHROPIC_API_KEY` is used when the provider is `anthropic`. `OPENAI_API_KEY` is used for **every other provider**, including `openrouter`. `OPENROUTER_API_KEY` is not read as a key.
 - If the chat says "check your key with /provider", the provider answered with an authentication error (401 or 403).
 - The test call checks that the model replies. It does not check that the model can call tools.
 
+## Connect several providers
+
+You can keep a key for every provider at the same time. Each key is saved on its own (in the Keychain, the Linux keyring or DPAPI), so switching provider never overwrites another key.
+
+1. In the chat, type `/connect`. You see every provider with `●` (connected) or `○` (not connected), and where the key comes from: a saved key, or an environment variable.
+2. Choose a provider. If it has no key, Amness asks for one (hidden). If it has one, you can use it now, replace the key, or remove the saved key.
+3. The menu stays open, so you can add the next provider. Press `Esc` when you are done.
+4. Switch between them with `/provider`. Amness uses the saved key of the provider you pick, and asks only if there is none.
+
+An environment variable wins over a saved key for its own provider: `ANTHROPIC_API_KEY` for Claude, `OPENROUTER_API_KEY` (or `OPENAI_API_KEY`) for OpenRouter, and `OPENAI_API_KEY` for the others. Environment keys are never saved. **Limit:** a key from an older version moves to the provider that was active when you first start the new version, so check `/connect` once.
+
 ## Switch provider, model or effort
 
 **How to switch in the chat.**
@@ -152,7 +163,8 @@ In `auto` mode Amness fixes these cases for you:
 
 | Variable | What it does |
 |---|---|
-| `OPENAI_API_KEY` | The key for every provider except `anthropic`. |
+| `OPENAI_API_KEY` | The key for every provider except `anthropic` (and for OpenRouter when `OPENROUTER_API_KEY` is not set). |
+| `OPENROUTER_API_KEY` | The key for `openrouter`. |
 | `ANTHROPIC_API_KEY` | The key for `anthropic`. |
 | `OPENAI_BASE_URL` | The base URL for non-Claude providers. |
 | `AMNESS_MODEL` | The model id. It wins over the saved config. |

package/docs/safety.md

@@ -35,7 +35,20 @@ Notes about the provider:
 
 ## Know where your keys are stored
 
-- Keys come from the environment variables `OPENAI_API_KEY` and `ANTHROPIC_API_KEY`, or from `~/.amness/config.json`. Amness writes that file with mode 600. Only your user can read it.
+- Keys come from the environment variables `OPENAI_API_KEY` and `ANTHROPIC_API_KEY`, or from the key store. A key from an environment variable is used for that run and is never saved.
+- The key store depends on your system. The prompt tells you which one is in use.
+
+  | System | Where the key is saved | Protection |
+  |---|---|---|
+  | macOS | the login Keychain (service `amness`, one entry per provider, for example `apiKey-anthropic`) | encrypted by macOS |
+  | Linux with a desktop keyring | the secret service, through `secret-tool` | encrypted by the keyring |
+  | Windows | `~/.amness/secrets/apiKey-<provider>.dpapi` | encrypted with DPAPI: only your Windows user on this machine can decrypt it |
+  | Docker or Kubernetes | nowhere: kept in memory for that run | nothing is written to the image or a volume. Pass the key as an environment variable or a secret file (`ANTHROPIC_API_KEY_FILE=/run/secrets/anthropic`). |
+  | Linux server with no keyring | `~/.amness/config.json` with mode 600 | **not encrypted**, only your user can read it. Prefer an environment variable or your platform's secret manager. |
+
+- Amness passes the key to the system tool on standard input, never on the command line, where other users could see it in the process list.
+- A key that an older version saved in `config.json` is moved to the key store the next time you start Amness, and removed from the file once the move is checked.
+- To remove a saved key, open the settings and clear "API key", or on macOS run `security delete-generic-password -s amness -a apiKey-anthropic` (use `apiKey-openai` and so on for the others).
 - The environment variable wins over the saved key.
 - Amness does not write keys into a project folder.
 - Commands the agent runs do not get your Amness keys. The variables `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `TAVILY_API_KEY`, `OPENROUTER_API_KEY` and `AMNESS_OTLP` are removed from their environment (`childEnv` in `src/secrets.ts`).

package/docs/usage.md

@@ -92,7 +92,9 @@ amness --inline
 - Set the environment variable `AMNESS_UI=inline`.
 - Type `/ui inline` in the chat. `/ui full` goes back.
 
-**What you see.** Each tool call is one line, for example `● Read(src/a.ts)`, with a short result. `Ctrl+O` shows everything. Each edit shows a diff with line numbers. A live todo list, a status line (elapsed time and tokens) and a footer (permission mode, git branch, model, context gauge, estimated cost) stay visible. Tool bullets have a colour for each kind and turn red when the tool fails. With `NO_COLOR=1`, or when the output is piped, Amness uses plain text.
+**What you see.** Each tool call is one line, for example `● Read(src/a.ts)`, with a short result. `Ctrl+O` shows everything. Each edit shows a diff with line numbers.
+
+In the full-screen view, the **most recent** tool call stays open. A shell command shows `$ the command` and its output, up to 20 lines. When the next tool call starts, the earlier one folds to one line again. Click a line to open or close it yourself. If the output is longer than 20 lines, the card says how many are hidden. Type `/output` to read all of it in a pager (`/output 2` for the call before). A lone file read does not open itself. A live todo list, a status line (elapsed time and tokens) and a footer (permission mode, git branch, model, context gauge, estimated cost) stay visible. Tool bullets have a colour for each kind and turn red when the tool fails. With `NO_COLOR=1`, or when the output is piped, Amness uses plain text.
 
 ## Type in the input box
 
@@ -204,14 +206,36 @@ The file is attached to your message. In the transcript you see a note such as "
 
 ## Slash commands
 
-Type `/` in the input box to see the popup. `/help` lists the commands in the chat.
+Type `/` in the input box to see the popup. It lists **every** command, grouped under titles: Conversation, Model & keys, Safety, Quality, Project, Extend and Interface. Your own and plugin commands go under Custom.
+
+- Move with `↑` and `↓`. `PgUp` and `PgDn` jump eight at a time. The top of the box shows where you are, for example `12/60 ↓`.
+- Keep typing to filter. The titles disappear and the best match comes first.
+- `Tab` or `Enter` picks the command. `Esc` closes the popup.
+
+**Completion after the command.** Type a command and a space, and the popup lists the values it accepts, each with a short explanation:
+
+| You type | The popup shows |
+|---|---|
+| `/mode ` | `default`, `acceptEdits`, `plan`, `yolo` |
+| `/model ` | the known models with their hints, then every other model your provider reports (OpenAI, Claude, OpenRouter, Ollama and other OpenAI-compatible servers). Amness asks the provider once and remembers the answer for ten minutes. If the provider cannot be reached, you still see the known models. |
+| `/models ` | the roles (`explore`, `compact`, `review`, `cheap`, `fallback`), then the models |
+| `/effort `, `/api `, `/theme `, `/shell `, `/strategy `, `/prompt `, `/sandbox `, `/diffmode `, `/ui ` | their choices |
+| `/keepwarm `, `/lsp `, `/approver `, `/reviewer `, `/router `, `/mouse ` | `on`, `off` |
+| `/branch ` | your turns, numbered, with the first words of each |
+| `/memory `, `/permissions `, `/copy `, `/pr `, `/report `, `/mcp `, `/budget `, `/check `, `/verify ` | their sub-commands or common values |
+
+`Enter` picks the highlighted value and runs the command. For commands that take free text or a number (`/model`, `/budget`, `/check`, `/verify`), `Enter` runs exactly what you typed unless you moved into the list with the arrow keys, so a model id that is not in the list still works. `Tab` always completes. Commands that need more words (`/memory add`, `/permissions allow`) add a space and wait.
+
+`/help` lists the commands in the chat.
 
 ### Models and providers
 
 | Command | What it does |
 |---|---|
 | `/model [id]` | Switches the model. Without an id, it opens a picker. |
-| `/provider` | Switches the provider (OpenAI, Claude, OpenRouter, Ollama, custom). The choice is saved. |
+| `/output [n]` (alias `/last`) | Shows the full output of the last tool call in a pager. `/output 2` shows the one before. |
+| `/provider` | Switches the provider (OpenAI, Claude, OpenRouter, Ollama, custom). The list shows `●` for providers that are connected and `○` for those that are not. Each provider uses its own saved key. |
+| `/connect` (alias `/keys`) | Opens the key manager. See which providers are connected, add a key for any of them, replace or remove one, or switch to it. Add as many as you want. |
 | `/effort [level]` | Sets the reasoning effort. Levels depend on the model. `default` lets the model decide. |
 | `/api [auto\|responses\|chat]` | Chooses the OpenAI wire API. It does not apply to Claude. |
 | `/models [role model]` | Uses another model for `explore`, `compact`, `review` or `cheap`. `/models fallback <model>` sets a model for when the main one is overloaded or not reachable. Use `off` to clear a role. |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "amness",
-  "version": "0.1.0",
+  "version": "0.1.3",
   "type": "module",
   "description": "A terminal coding agent for any model: streaming chat TUI, permission modes, plan mode, undo, memory, MCP, with a small reliability harness underneath. OpenAI (Responses API), Claude, and local models.",
   "license": "MIT",
@@ -52,19 +52,13 @@
     "test:pty": "python3 test/pty/tui.py",
     "prepack": "npm run build"
   },
+  "optionalDependencies": {
+    "@anthropic-ai/sdk": "^1.3.0"
+  },
   "devDependencies": {
-    "@anthropic-ai/sdk": "^1.3.0",
     "@types/node": "^22.0.0",
     "typescript": "^5.8.0"
   },
-  "peerDependencies": {
-    "@anthropic-ai/sdk": "^1.3.0"
-  },
-  "peerDependenciesMeta": {
-    "@anthropic-ai/sdk": {
-      "optional": true
-    }
-  },
   "files": [
     "dist",
     "docs",