@modudraft/mcp 0.23.3 → 0.24.0

package/README.md

@@ -90,12 +90,28 @@ Every MCP tool is listed by default. If that is too much context for your client
 | `analysis` | Validate, diff, cross-diagram suggestions, clone | `validate_diagram`, `diff_diagram`, `find_shared_services`, `suggest_connections`, `clone_diagram`, `suggest_architecture` |
 | `project` | Watch a repo for drift, sync a diagram with the repo, generate sequences from the API | `watch_project`, `check_project_drift`, `unwatch_project`, `sync_project`, `generate_sequence_from_api` |
 | `history` | Server version history (snapshots) and the edit log | `snapshot_diagram`, `list_snapshots`, `restore_snapshot`, `get_edit_log` |
+| `security` | Threat model: trust zones, data classes, STRIDE threats, the Markdown report | `get_security_view`, `suggest_threats`, `add_threat`, `update_threat`, `delete_threat`, `set_trust_zone`, `set_data_class`, `export_threat_model` |
 
 - `MODUDRAFT_TOOLSETS="core"` lists the core tools only; `"sequences,api"` lists core plus those groups; `"all"` (or unset) lists everything.
 - A value made only of unknown names keeps the default (everything) and prints a warning on stderr.
 - `load_toolset(name)` adds a group at runtime and sends a tools-changed notification. If your client doesn't refresh tools after `load_toolset`, set `MODUDRAFT_TOOLSETS` instead.
 - Calling any tool by name always works, even when it isn't listed: the server loads its group and runs it.
 
+### Upgrading to 0.24.0
+
+- New `security` toolset for threat modeling the Architecture tab (file format 2.12.0). Deploy the web app and API first: a diagram with threats needs a web app that reads file format 2.12.0.
+  - `set_trust_zone` sets a boundary's trust zone (`internet`, `dmz`, `internal`, `restricted`, `third-party`). Items inside take its zone.
+  - `set_data_class` sets the data class (`public`, `internal`, `pii`, `credentials`, `payment`) of an arrow or a data store, plus `encrypted` (arrows, in transit) or `encrypted_at_rest` (data stores). These fields are set only here; `update_edge` and `update_node` are unchanged.
+  - `get_security_view` returns the zones, each component's tool, archetype and zone, each arrow's protocol, data class and encryption, and the threats.
+  - `suggest_threats` runs the same STRIDE rules as the web app's "Suggest threats" and adds the new suggestions. Running it again never adds a duplicate and never brings back a threat marked `not-applicable`.
+  - `add_threat`, `update_threat` and `delete_threat` edit the threat list. Threats the agent adds are marked `source: agent`.
+  - `export_threat_model` returns the Markdown report, or writes it to a `.md` file under your project folders with `path`.
+- The workflow the tools describe: set zones and data classes, run `suggest_threats`, call `get_security_view` and add threats specific to each component's tool and archetype (the rules are generic and don't know a technology's own weaknesses), then review each suggestion and fill in mitigations and owners, and export the report.
+- Threat edits are sent as targeted threat ops, so they don't overwrite concurrent canvas edits in the web app.
+- `clone_diagram` now points a cloned diagram's threats at the cloned items. Clients below 0.24.0 keep threats on their writes, but their `clone_diagram` copies threats with the original target ids, so the copy's threats show "target removed". Upgrade before cloning a diagram that has a threat model.
+- Threat text is capped: title 200 characters, description and mitigation 2000, owner 100. `add_threat` and `update_threat` refuse longer text with a message naming the field.
+- With more than one Architecture tab, the first tab that gets a threat holds the model; the web app, the report and these tools all follow it.
+
 ### Upgrading to 0.23.3
 
 - The "MODUDRAFT_API_KEY is not set" message now links to the same API keys page as the expired and revoked key messages (`/account?section=api-keys`) instead of `/dashboard`.
@@ -158,7 +174,7 @@ Every MCP tool is listed by default. If that is too much context for your client
 Besides tools, the server offers MCP resources and prompts, for clients that support them.
 
 - **Resources:** each diagram is readable as plain-text line art at `modudraft://diagram/<id>` (the architecture tab, the same output as `render_diagram_text`). `resources/list` returns your 50 most recently edited diagrams; any diagram id you can reach can be read by URI.
-- **Prompts:** `map_repo` (optional `path`) scans a repository and imports what it finds; `check_drift` (`diagram_id`) checks a diagram against the project code; `review_diagram` (`diagram_id`) validates a diagram and lists its issues. Each is one message that names the toolset to load.
+- **Prompts:** `map_repo` (optional `path`) scans a repository and imports what it finds; `check_drift` (`diagram_id`) checks a diagram against the project code; `review_diagram` (`diagram_id`) validates a diagram and lists its issues; `threat_model` (`diagram_id`) builds a STRIDE threat model and exports the report. Each is one message that names the toolset to load.
 
 ### Upgrading to 0.17.0
 
@@ -368,6 +384,22 @@ The old `request_body`, `request_format`, `body`, `body_format` and top-level `r
 | `get_org_context` | Get cached org context - existing tools, protocols, naming patterns |
 | `refresh_org_context` | Force a refresh of the org context cache |
 
+### Threat model
+*(toolset: `security`)*
+
+Threats live on the Architecture tab (one threat model per diagram). Run `suggest_threats` first, then add threats specific to each component's tool and archetype, then fill in mitigations.
+
+| Tool | Description |
+|---|---|
+| `get_security_view` | Boundaries with their trust zones; each component's tool, archetype, effective zone and (data stores) data class and encryption at rest; each arrow's protocol, data class and encryption in transit; the threats (`targetRemoved: true` when the item was deleted) |
+| `suggest_threats` | Run the Modudraft STRIDE rules and add the new suggestions (status `suggested`). Returns the count and titles. Safe to re-run: no duplicates, and a `not-applicable` threat never comes back |
+| `add_threat` | Add a threat (`source: agent`): `title`, `stride` (S T R I D E), `target` (`{ kind: "node" \| "edge" \| "boundary" \| "diagram", id }`), optional `severity`, `status`, `description`, `mitigation`, `owner` |
+| `update_threat` | Change any field of a threat; an empty string clears `description`, `mitigation` or `owner` |
+| `delete_threat` | Delete a threat (for a rule suggestion that doesn't apply, prefer status `not-applicable`) |
+| `set_trust_zone` | Set or clear a boundary's trust zone: `internet`, `dmz`, `internal`, `restricted`, `third-party` |
+| `set_data_class` | Set `data_class` on an arrow or a data store, plus `encrypted` (arrows) or `encrypted_at_rest` (data stores) |
+| `export_threat_model` | The Markdown threat-model report (same as the web app's download), or write it to a `.md` file with `path` |
+
 ---
 
 ## Example prompts