fullcourtdefense-cli 1.7.26 → 1.7.28

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (3) hide show
  1. package/dist/index.js +232 -232
  2. package/dist/version.json +1 -1
  3. package/package.json +1 -1
package/dist/index.js CHANGED
@@ -104,238 +104,238 @@ function parseArgs(argv) {
104
104
  return { command, flags };
105
105
  }
106
106
  function printHelp() {
107
- console.log(`
108
- \x1b[1m\x1b[36mFullCourtDefense CLI\x1b[0m v${VERSION}
109
- Security scanning for AI agents from your terminal.
110
-
111
- \x1b[1mUsage:\x1b[0m
112
- fullcourtdefense <command> [options]
113
-
114
- \x1b[1mCommands:\x1b[0m
115
- help Show this onboarding guide and command reference.
116
- doctor First step. Checks outbound HTTPS access to FullCourtDefense.
117
- login Enrolls this machine with a fleet enrollment token and saves
118
- per-machine Shield credentials to ~/.fullcourtdefense.yml.
119
- configure Legacy/manual setup: saves org API key, Organization ID,
120
- Shield ID, and Shield key to ~/.fullcourtdefense.yml.
121
- setup Same as configure — manual fallback when not using fleet login.
122
- install-all One-click: protect existing MCP servers in every AI client and
123
- install available runtime hooks for built-in actions. No folder path
124
- or --mcp-command needed. Use --hooks false to skip runtime hooks.
125
- install Alias for install-all.
126
- scan Runs the scan. Use --local for inside-organization scans.
127
- discover Finds MCP servers, local secrets, agent rules/skills, and machine
128
- blast radius on this laptop. Use --surface all --upload for AI Fleet.
129
- Use --schedule logon to upload inventory/posture at machine login.
130
- agent-ci CI/CD gate for agents and tools only. Fails builds on risky MCP,
131
- agent instruction, gateway, policy, or secret drift. Not a bot scan.
132
- install-cursor-hook
133
- Installs a Cursor hook so EVERY agent action on this machine is
134
- checked against your org's Action Policies — in any repo/folder.
135
- Prompts are scanned by your Shield; shell/MCP/file/read actions
136
- are allowed, blocked, or paused for human approval per policy.
137
- uninstall-cursor-hook
138
- Removes the FullCourtDefense Cursor hook entries.
139
- install-claude-hook
140
- Installs Claude-format hooks in ~/.claude/settings.json — ONE file
141
- enforces Action Policies + Local Safety rules in Claude Code,
142
- VS Code Copilot agent mode, and GitHub Copilot CLI.
143
- uninstall-claude-hook
144
- Removes the FullCourtDefense Claude-format hook entries.
145
- protect-all Auto-detect EVERY MCP server already configured in EVERY client
146
- (Cursor, Claude Code/Desktop, Codex, Gemini, Windsurf, VS Code)
147
- and wrap each through the Shield proxy — no --mcp-command needed.
148
- Enforces your org Action Policies on every tool. --dry-run to preview.
149
- Per-tool: --clients cursor (or claude-code, claude-desktop, codex,
150
- gemini-cli, windsurf, vscode; comma-separated). Default: all tools.
151
- unprotect-all
152
- Reverse protect-all: restore every server to its original command.
153
- Also supports --clients to target one tool, and --dry-run.
154
- auto-protect
155
- Self-healing: schedule protect-all to re-run automatically so any
156
- server that loses the proxy is re-wrapped. --interval-minutes 60
157
- (default) or --on-logon true. --uninstall true / --status true.
158
- install-mcp-gateway
159
- Install MCP gateway into selected clients (--clients all for every tool).
160
- install-cursor-mcp-gateway
161
- Cursor only — adds gateway to mcp.json.
162
- install-claude-code-mcp-gateway
163
- Claude Code only — uses claude mcp add or .mcp.json.
164
- install-claude-desktop-mcp-gateway
165
- Claude Desktop only.
166
- install-codex-mcp-gateway
167
- Codex only — writes ~/.codex/config.toml.
168
- install-gemini-mcp-gateway
169
- Gemini CLI only — writes settings.json mcpServers.
170
- install-windsurf-mcp-gateway
171
- Windsurf only.
172
- install-vscode-mcp-gateway
173
- VS Code only — .vscode/mcp.json or user settings.
174
- mcp-gateway
175
- Runs a local MCP gateway that checks AgentGuard runtime/action
176
- policies before forwarding tool calls to a real MCP server.
177
- hook Internal: invoked by Cursor per agent event (reads stdin JSON,
178
- returns an allow/deny verdict). Not run by hand.
179
- credits Shows hosted scan credits for CI/CD API-key scans.
180
- init Creates a starter .fullcourtdefense.yml config file.
181
-
182
- \x1b[1mNew User Process:\x1b[0m
183
- 1. In the web app, an org admin creates a fleet enrollment token:
184
- AI Fleet → Settings → Fleet enrollment token.
185
-
186
- 2. Check outbound connectivity from the customer machine:
187
- fullcourtdefense doctor
188
- This confirms the machine can reach https://api.fullcourtdefense.ai over HTTPS.
189
-
190
- 3. Enroll this developer machine (one time):
191
- fullcourtdefense login --token <fleet-enrollment-token>
192
- Or set FCD_ENROLL_TOKEN and run: fullcourtdefense login
193
- Saves per-machine Shield credentials to ~/.fullcourtdefense.yml —
194
- install, discover --upload, scan, gateway, and hooks use it automatically.
195
-
196
- 4. One-click protect every AI client on this machine with Action Policies:
197
- fullcourtdefense install-all
198
- Verify installs:
199
- fullcourtdefense protect-all --dry-run true
200
- Or install one client: fullcourtdefense install-cursor-mcp-gateway ...
201
-
202
- 5. Run a guided local scan:
203
- fullcourtdefense scan --local
204
- The CLI asks whether to scan endpoint, mcp, or rag.
205
- To open the localhost web scan UI first:
206
- fullcourtdefense scan --local --open-ui
207
-
208
- 6. Review results:
209
- Use --format summary for CI, --format table for terminal view,
210
- --format report for evidence, --format full-report for every row,
211
- or --format json for raw output.
212
- With a Shield key configured, reports are saved to the web Reports page
213
- and the browser opens to the saved report. Use --no-open to suppress.
214
-
215
- \x1b[1mLocal Destinations:\x1b[0m
216
- endpoint Internal HTTP URL, for example http://agent.local/chat.
217
- The CLI sends attack prompts to this endpoint and scans responses.
218
-
219
- mcp Either stdio command + args, or an already-running HTTP/SSE MCP URL.
220
- Stdio example: node ./mcp-server.js
221
- HTTP/HTTPS examples: http://mcp.internal/mcp or https://internal.company.com/mcp
222
- Legacy SSE example: https://internal.company.com/sse
223
- Secured HTTP MCP supports bearer, basic, and API-key auth.
224
-
225
- rag File/directory path or live RAG HTTP endpoint.
226
- Use --rag-path for local documents, or --rag-url for a RAG service.
227
-
228
- \x1b[1mRecommended Commands:\x1b[0m
229
- fullcourtdefense doctor
230
- fullcourtdefense configure
231
- fullcourtdefense agent-ci --fail-on high --require-gateway --format summary
232
- fullcourtdefense agent-ci --fail-on high --format sarif --upload
233
- fullcourtdefense scan --local
234
- fullcourtdefense scan --local --type mcp --mcp-command node --mcp-args ./server.js --mcp-tool all --mode full --format report
235
- fullcourtdefense scan --local --type mcp --mcp-url https://internal.company.com/mcp --mcp-auth-type bearer --mcp-token TOKEN --mcp-tool all --mode full --format report
236
- fullcourtdefense scan --local --type rag --rag-path ./docs --format report
237
- fullcourtdefense scan --local --type rag --rag-url http://rag.internal/chat --method POST --request-format custom --input-field message --output-field answer --mode full --format report
238
- fullcourtdefense scan --local --type endpoint --endpoint http://agent.local/chat --method POST --mode full --format report
239
-
240
- \x1b[1mScan Options:\x1b[0m
241
- --local Run scan from this machine, inside your network
242
- --open-ui Open localhost browser scan UI with all options
243
- --type <type> Local scan type: endpoint, mcp, rag
244
- --api-key <key> API key (or set FULLCOURTDEFENSE_API_KEY env var)
245
- --api-url <url> API base URL (default: https://api.fullcourtdefense.ai)
246
- --endpoint <url> Agent API endpoint to scan
247
- --mcp-url <url> Already-running HTTP MCP server URL
248
- --mcp-transport <t> MCP transport: stdio, http, or sse
249
- --mcp-command <cmd> MCP server command, for example node or npx.cmd
250
- --mcp-args <args> MCP server args/path, for example ./server.js
251
- --mcp-tool <tool> MCP tool to test, or all
252
- --server-name <name> MCP gateway server name for client installers
253
- --clients <list> Gateway clients: all, auto, cursor, claude-code,
254
- claude-desktop, codex, gemini-cli, windsurf, vscode
255
- --developer-name <name> Runtime identity for user-scoped policies (auto-detected if omitted)
256
- --agent-name <name> Runtime agent name (auto-generated from identity + client if omitted)
257
- --scope <scope> Claude Code MCP scope: local, project, or user
258
- --claude-code-scope <s> Claude Code scope for install-mcp-gateway
259
- --cursor-project <bool> Cursor project config for install-mcp-gateway
260
- --config-path <path> Claude Desktop config override path
261
- --mcp-auth-type <type> HTTP MCP auth: none, bearer, basic, api-key
262
- --mcp-token <token> HTTP MCP bearer token
263
- --mcp-username <user> HTTP MCP basic username
264
- --mcp-password <pass> HTTP MCP basic password
265
- --mcp-api-key-header <h> HTTP MCP API-key header name
266
- --mcp-api-key <key> HTTP MCP API key value
267
- --rag-path <path> RAG file or directory to scan
268
- --rag-url <url> Live RAG HTTP endpoint to scan
269
- --shield-id <id> Shield ID for local outbound Shield verdicts
270
- --shield-key <key> Optional Shield key for locked Shields
271
- --fail-on <severity> agent-ci fail threshold: critical, high, medium, low
272
- --require-gateway agent-ci requires risky MCP servers to use FCD gateway
273
- --require-approval-for agent-ci risky controls list, e.g. payment,delete,shell,secrets
274
- --require-upload agent-ci fails if evidence cannot upload to AgentGuard UI
275
- --changed-only agent-ci limits findings to changed files
276
- --changed-files <path> file or comma/newline list of changed files for agent-ci
277
- --method <GET|POST> Local endpoint method
278
- --request-format <fmt> Local endpoint request format: custom or openai
279
- --auth-type <type> none, bearer, basic, api-key
280
- --description <text> Agent description
281
- --system-prompt <text> System prompt (text or path to file)
282
- --mode <sync|async> Hosted mode, or local mode: quick, full, targeted, deep
283
- --categories <list> Comma-separated attack categories
284
- --attack-count <n> Number of attacks to run
285
- --surface <list> Discover surfaces: mcp, secrets, agent-files, posture, all
286
- --fail-threshold <n> Exit code 1 if score below n (default: 0)
287
- --format <fmt> Output: table, json, summary, report, full-report (default: table)
288
- --open Open browser scan UI and saved local scan report
289
- --open-ui Open only the browser scan UI before local execution
290
- --no-open Do not open the browser during/after local scans
291
- --webhook-format <fmt> Hosted webhook format, or legacy local request format
292
- --config <path> Path to .fullcourtdefense.yml config file
293
-
294
- \x1b[1mExamples:\x1b[0m
295
- $ fullcourtdefense scan --endpoint https://my-bot.com/chat --description "Support bot"
296
- $ fullcourtdefense help
297
- $ fullcourtdefense scan --local
298
- $ fullcourtdefense scan --local --open-ui
299
- $ fullcourtdefense scan --local --type endpoint --mode full --endpoint http://internal-agent/chat
300
- $ fullcourtdefense scan --local --type mcp --mcp-command node --mcp-args ./server.js --mcp-tool all --mode full
301
- $ fullcourtdefense scan --local --type mcp --mcp-url https://internal.company.com/mcp
302
- $ fullcourtdefense scan --local --type mcp --mcp-url https://internal.company.com/mcp --mcp-tool all --mode full
303
- $ fullcourtdefense scan --local --type mcp --mcp-transport sse --mcp-url https://internal.company.com/sse --mcp-tool all --mode full
304
- $ fullcourtdefense scan --local --type mcp --mcp-command node --mcp-args ./server.js --mcp-tool all --mode full --format report
305
- $ fullcourtdefense discover
306
- $ fullcourtdefense discover --surface mcp
307
- $ fullcourtdefense discover --surface all --upload
308
- $ fullcourtdefense discover --surface secrets
309
- $ fullcourtdefense discover --surface agent-files
310
- $ fullcourtdefense discover --deep --upload --silent --user-email you@company.com
311
- $ fullcourtdefense discover --type mcp --json
312
- $ fullcourtdefense install-cursor-hook
313
- $ fullcourtdefense install-cursor-hook --shadow true # monitor only
314
- $ fullcourtdefense install-cursor-hook --events prompt,shell,mcp,file
315
- $ fullcourtdefense install-cursor-hook --project true # this repo only
316
- $ fullcourtdefense uninstall-cursor-hook
317
- $ fullcourtdefense install-claude-hook # Claude Code + VS Code + Copilot CLI
318
- $ fullcourtdefense install-claude-hook --events tools,prompt # also scan prompts
319
- $ fullcourtdefense uninstall-claude-hook
320
- $ fullcourtdefense install-mcp-gateway --clients auto --mcp-command npm --mcp-args "run mcp"
321
- $ fullcourtdefense install-mcp-gateway --clients all --mcp-command npm --mcp-args "run mcp" --upload
322
- $ fullcourtdefense install-cursor-mcp-gateway --project true --mcp-command npm --mcp-args "run mcp"
323
- $ fullcourtdefense install-claude-code-mcp-gateway --scope local --mcp-command npm --mcp-args "run mcp"
324
- $ fullcourtdefense install-claude-code-mcp-gateway --scope project --mcp-command npm --mcp-args "run mcp"
325
- $ fullcourtdefense install-claude-desktop-mcp-gateway --mcp-command npm --mcp-args "run mcp"
326
- $ fullcourtdefense mcp-gateway --mcp-command npm --mcp-args "run mcp"
327
- $ fullcourtdefense login --token <fleet-enrollment-token>
328
- $ fullcourtdefense install-all
329
- $ fullcourtdefense install-all --auto-protect true --upload
330
- $ fullcourtdefense discover --surface all --upload --schedule logon
331
- $ fullcourtdefense install-mcp-gateway --clients cursor,codex --mcp-command npm --mcp-args "run mcp"
332
- $ fullcourtdefense install-codex-mcp-gateway --mcp-command npm --mcp-args "run mcp"
333
- $ fullcourtdefense scan --system-prompt ./prompts/system.md --fail-threshold 80
334
- $ fullcourtdefense scan --config .fullcourtdefense.yml --format json
335
- $ fullcourtdefense credits
336
- $ fullcourtdefense init
337
-
338
- \x1b[2mDocs: https://fullcourtdefense.ai/docs/cli\x1b[0m
107
+ console.log(`
108
+ \x1b[1m\x1b[36mFullCourtDefense CLI\x1b[0m v${VERSION}
109
+ Security scanning for AI agents from your terminal.
110
+
111
+ \x1b[1mUsage:\x1b[0m
112
+ fullcourtdefense <command> [options]
113
+
114
+ \x1b[1mCommands:\x1b[0m
115
+ help Show this onboarding guide and command reference.
116
+ doctor First step. Checks outbound HTTPS access to FullCourtDefense.
117
+ login Enrolls this machine with a fleet enrollment token and saves
118
+ per-machine Shield credentials to ~/.fullcourtdefense.yml.
119
+ configure Legacy/manual setup: saves org API key, Organization ID,
120
+ Shield ID, and Shield key to ~/.fullcourtdefense.yml.
121
+ setup Same as configure — manual fallback when not using fleet login.
122
+ install-all One-click: protect existing MCP servers in every AI client and
123
+ install available runtime hooks for built-in actions. No folder path
124
+ or --mcp-command needed. Use --hooks false to skip runtime hooks.
125
+ install Alias for install-all.
126
+ scan Runs the scan. Use --local for inside-organization scans.
127
+ discover Finds MCP servers, local secrets, agent rules/skills, and machine
128
+ blast radius on this laptop. Use --surface all --upload for AI Fleet.
129
+ Use --schedule logon to upload inventory/posture at machine login.
130
+ agent-ci CI/CD gate for agents and tools only. Fails builds on risky MCP,
131
+ agent instruction, gateway, policy, or secret drift. Not a bot scan.
132
+ install-cursor-hook
133
+ Installs a Cursor hook so EVERY agent action on this machine is
134
+ checked against your org's Action Policies — in any repo/folder.
135
+ Prompts are scanned by your Shield; shell/MCP/file/read actions
136
+ are allowed, blocked, or paused for human approval per policy.
137
+ uninstall-cursor-hook
138
+ Removes the FullCourtDefense Cursor hook entries.
139
+ install-claude-hook
140
+ Installs Claude-format hooks in ~/.claude/settings.json — ONE file
141
+ enforces Action Policies + Local Safety rules in Claude Code,
142
+ VS Code Copilot agent mode, and GitHub Copilot CLI.
143
+ uninstall-claude-hook
144
+ Removes the FullCourtDefense Claude-format hook entries.
145
+ protect-all Auto-detect EVERY MCP server already configured in EVERY client
146
+ (Cursor, Claude Code/Desktop, Codex, Gemini, Windsurf, VS Code)
147
+ and wrap each through the Shield proxy — no --mcp-command needed.
148
+ Enforces your org Action Policies on every tool. --dry-run to preview.
149
+ Per-tool: --clients cursor (or claude-code, claude-desktop, codex,
150
+ gemini-cli, windsurf, vscode; comma-separated). Default: all tools.
151
+ unprotect-all
152
+ Reverse protect-all: restore every server to its original command.
153
+ Also supports --clients to target one tool, and --dry-run.
154
+ auto-protect
155
+ Self-healing: schedule protect-all to re-run automatically so any
156
+ server that loses the proxy is re-wrapped. --interval-minutes 60
157
+ (default) or --on-logon true. --uninstall true / --status true.
158
+ install-mcp-gateway
159
+ Install MCP gateway into selected clients (--clients all for every tool).
160
+ install-cursor-mcp-gateway
161
+ Cursor only — adds gateway to mcp.json.
162
+ install-claude-code-mcp-gateway
163
+ Claude Code only — uses claude mcp add or .mcp.json.
164
+ install-claude-desktop-mcp-gateway
165
+ Claude Desktop only.
166
+ install-codex-mcp-gateway
167
+ Codex only — writes ~/.codex/config.toml.
168
+ install-gemini-mcp-gateway
169
+ Gemini CLI only — writes settings.json mcpServers.
170
+ install-windsurf-mcp-gateway
171
+ Windsurf only.
172
+ install-vscode-mcp-gateway
173
+ VS Code only — .vscode/mcp.json or user settings.
174
+ mcp-gateway
175
+ Runs a local MCP gateway that checks AgentGuard runtime/action
176
+ policies before forwarding tool calls to a real MCP server.
177
+ hook Internal: invoked by Cursor per agent event (reads stdin JSON,
178
+ returns an allow/deny verdict). Not run by hand.
179
+ credits Shows hosted scan credits for CI/CD API-key scans.
180
+ init Creates a starter .fullcourtdefense.yml config file.
181
+
182
+ \x1b[1mNew User Process:\x1b[0m
183
+ 1. In the web app, an org admin creates a fleet enrollment token:
184
+ AI Fleet → Settings → Fleet enrollment token.
185
+
186
+ 2. Check outbound connectivity from the customer machine:
187
+ fullcourtdefense doctor
188
+ This confirms the machine can reach https://api.fullcourtdefense.ai over HTTPS.
189
+
190
+ 3. Enroll this developer machine (one time):
191
+ fullcourtdefense login --token <fleet-enrollment-token>
192
+ Or set FCD_ENROLL_TOKEN and run: fullcourtdefense login
193
+ Saves per-machine Shield credentials to ~/.fullcourtdefense.yml —
194
+ install, discover --upload, scan, gateway, and hooks use it automatically.
195
+
196
+ 4. One-click protect every AI client on this machine with Action Policies:
197
+ fullcourtdefense install-all
198
+ Verify installs:
199
+ fullcourtdefense protect-all --dry-run true
200
+ Or install one client: fullcourtdefense install-cursor-mcp-gateway ...
201
+
202
+ 5. Run a guided local scan:
203
+ fullcourtdefense scan --local
204
+ The CLI asks whether to scan endpoint, mcp, or rag.
205
+ To open the localhost web scan UI first:
206
+ fullcourtdefense scan --local --open-ui
207
+
208
+ 6. Review results:
209
+ Use --format summary for CI, --format table for terminal view,
210
+ --format report for evidence, --format full-report for every row,
211
+ or --format json for raw output.
212
+ With a Shield key configured, reports are saved to the web Reports page
213
+ and the browser opens to the saved report. Use --no-open to suppress.
214
+
215
+ \x1b[1mLocal Destinations:\x1b[0m
216
+ endpoint Internal HTTP URL, for example http://agent.local/chat.
217
+ The CLI sends attack prompts to this endpoint and scans responses.
218
+
219
+ mcp Either stdio command + args, or an already-running HTTP/SSE MCP URL.
220
+ Stdio example: node ./mcp-server.js
221
+ HTTP/HTTPS examples: http://mcp.internal/mcp or https://internal.company.com/mcp
222
+ Legacy SSE example: https://internal.company.com/sse
223
+ Secured HTTP MCP supports bearer, basic, and API-key auth.
224
+
225
+ rag File/directory path or live RAG HTTP endpoint.
226
+ Use --rag-path for local documents, or --rag-url for a RAG service.
227
+
228
+ \x1b[1mRecommended Commands:\x1b[0m
229
+ fullcourtdefense doctor
230
+ fullcourtdefense configure
231
+ fullcourtdefense agent-ci --fail-on high --require-gateway --format summary
232
+ fullcourtdefense agent-ci --fail-on high --format sarif --upload
233
+ fullcourtdefense scan --local
234
+ fullcourtdefense scan --local --type mcp --mcp-command node --mcp-args ./server.js --mcp-tool all --mode full --format report
235
+ fullcourtdefense scan --local --type mcp --mcp-url https://internal.company.com/mcp --mcp-auth-type bearer --mcp-token TOKEN --mcp-tool all --mode full --format report
236
+ fullcourtdefense scan --local --type rag --rag-path ./docs --format report
237
+ fullcourtdefense scan --local --type rag --rag-url http://rag.internal/chat --method POST --request-format custom --input-field message --output-field answer --mode full --format report
238
+ fullcourtdefense scan --local --type endpoint --endpoint http://agent.local/chat --method POST --mode full --format report
239
+
240
+ \x1b[1mScan Options:\x1b[0m
241
+ --local Run scan from this machine, inside your network
242
+ --open-ui Open localhost browser scan UI with all options
243
+ --type <type> Local scan type: endpoint, mcp, rag
244
+ --api-key <key> API key (or set FULLCOURTDEFENSE_API_KEY env var)
245
+ --api-url <url> API base URL (default: https://api.fullcourtdefense.ai)
246
+ --endpoint <url> Agent API endpoint to scan
247
+ --mcp-url <url> Already-running HTTP MCP server URL
248
+ --mcp-transport <t> MCP transport: stdio, http, or sse
249
+ --mcp-command <cmd> MCP server command, for example node or npx.cmd
250
+ --mcp-args <args> MCP server args/path, for example ./server.js
251
+ --mcp-tool <tool> MCP tool to test, or all
252
+ --server-name <name> MCP gateway server name for client installers
253
+ --clients <list> Gateway clients: all, auto, cursor, claude-code,
254
+ claude-desktop, codex, gemini-cli, windsurf, vscode
255
+ --developer-name <name> Runtime identity for user-scoped policies (auto-detected if omitted)
256
+ --agent-name <name> Runtime agent name (auto-generated from identity + client if omitted)
257
+ --scope <scope> Claude Code MCP scope: local, project, or user
258
+ --claude-code-scope <s> Claude Code scope for install-mcp-gateway
259
+ --cursor-project <bool> Cursor project config for install-mcp-gateway
260
+ --config-path <path> Claude Desktop config override path
261
+ --mcp-auth-type <type> HTTP MCP auth: none, bearer, basic, api-key
262
+ --mcp-token <token> HTTP MCP bearer token
263
+ --mcp-username <user> HTTP MCP basic username
264
+ --mcp-password <pass> HTTP MCP basic password
265
+ --mcp-api-key-header <h> HTTP MCP API-key header name
266
+ --mcp-api-key <key> HTTP MCP API key value
267
+ --rag-path <path> RAG file or directory to scan
268
+ --rag-url <url> Live RAG HTTP endpoint to scan
269
+ --shield-id <id> Shield ID for local outbound Shield verdicts
270
+ --shield-key <key> Optional Shield key for locked Shields
271
+ --fail-on <severity> agent-ci fail threshold: critical, high, medium, low
272
+ --require-gateway agent-ci requires risky MCP servers to use FCD gateway
273
+ --require-approval-for agent-ci risky controls list, e.g. payment,delete,shell,secrets
274
+ --require-upload agent-ci fails if evidence cannot upload to AgentGuard UI
275
+ --changed-only agent-ci limits findings to changed files
276
+ --changed-files <path> file or comma/newline list of changed files for agent-ci
277
+ --method <GET|POST> Local endpoint method
278
+ --request-format <fmt> Local endpoint request format: custom or openai
279
+ --auth-type <type> none, bearer, basic, api-key
280
+ --description <text> Agent description
281
+ --system-prompt <text> System prompt (text or path to file)
282
+ --mode <sync|async> Hosted mode, or local mode: quick, full, targeted, deep
283
+ --categories <list> Comma-separated attack categories
284
+ --attack-count <n> Number of attacks to run
285
+ --surface <list> Discover surfaces: mcp, secrets, agent-files, posture, all
286
+ --fail-threshold <n> Exit code 1 if score below n (default: 0)
287
+ --format <fmt> Output: table, json, summary, report, full-report (default: table)
288
+ --open Open browser scan UI and saved local scan report
289
+ --open-ui Open only the browser scan UI before local execution
290
+ --no-open Do not open the browser during/after local scans
291
+ --webhook-format <fmt> Hosted webhook format, or legacy local request format
292
+ --config <path> Path to .fullcourtdefense.yml config file
293
+
294
+ \x1b[1mExamples:\x1b[0m
295
+ $ fullcourtdefense scan --endpoint https://my-bot.com/chat --description "Support bot"
296
+ $ fullcourtdefense help
297
+ $ fullcourtdefense scan --local
298
+ $ fullcourtdefense scan --local --open-ui
299
+ $ fullcourtdefense scan --local --type endpoint --mode full --endpoint http://internal-agent/chat
300
+ $ fullcourtdefense scan --local --type mcp --mcp-command node --mcp-args ./server.js --mcp-tool all --mode full
301
+ $ fullcourtdefense scan --local --type mcp --mcp-url https://internal.company.com/mcp
302
+ $ fullcourtdefense scan --local --type mcp --mcp-url https://internal.company.com/mcp --mcp-tool all --mode full
303
+ $ fullcourtdefense scan --local --type mcp --mcp-transport sse --mcp-url https://internal.company.com/sse --mcp-tool all --mode full
304
+ $ fullcourtdefense scan --local --type mcp --mcp-command node --mcp-args ./server.js --mcp-tool all --mode full --format report
305
+ $ fullcourtdefense discover
306
+ $ fullcourtdefense discover --surface mcp
307
+ $ fullcourtdefense discover --surface all --upload
308
+ $ fullcourtdefense discover --surface secrets
309
+ $ fullcourtdefense discover --surface agent-files
310
+ $ fullcourtdefense discover --deep --upload --silent --user-email you@company.com
311
+ $ fullcourtdefense discover --type mcp --json
312
+ $ fullcourtdefense install-cursor-hook
313
+ $ fullcourtdefense install-cursor-hook --shadow true # monitor only
314
+ $ fullcourtdefense install-cursor-hook --events prompt,shell,mcp,file
315
+ $ fullcourtdefense install-cursor-hook --project true # this repo only
316
+ $ fullcourtdefense uninstall-cursor-hook
317
+ $ fullcourtdefense install-claude-hook # Claude Code + VS Code + Copilot CLI
318
+ $ fullcourtdefense install-claude-hook --events tools,prompt # also scan prompts
319
+ $ fullcourtdefense uninstall-claude-hook
320
+ $ fullcourtdefense install-mcp-gateway --clients auto --mcp-command npm --mcp-args "run mcp"
321
+ $ fullcourtdefense install-mcp-gateway --clients all --mcp-command npm --mcp-args "run mcp" --upload
322
+ $ fullcourtdefense install-cursor-mcp-gateway --project true --mcp-command npm --mcp-args "run mcp"
323
+ $ fullcourtdefense install-claude-code-mcp-gateway --scope local --mcp-command npm --mcp-args "run mcp"
324
+ $ fullcourtdefense install-claude-code-mcp-gateway --scope project --mcp-command npm --mcp-args "run mcp"
325
+ $ fullcourtdefense install-claude-desktop-mcp-gateway --mcp-command npm --mcp-args "run mcp"
326
+ $ fullcourtdefense mcp-gateway --mcp-command npm --mcp-args "run mcp"
327
+ $ fullcourtdefense login --token <fleet-enrollment-token>
328
+ $ fullcourtdefense install-all
329
+ $ fullcourtdefense install-all --auto-protect true --upload
330
+ $ fullcourtdefense discover --surface all --upload --schedule logon
331
+ $ fullcourtdefense install-mcp-gateway --clients cursor,codex --mcp-command npm --mcp-args "run mcp"
332
+ $ fullcourtdefense install-codex-mcp-gateway --mcp-command npm --mcp-args "run mcp"
333
+ $ fullcourtdefense scan --system-prompt ./prompts/system.md --fail-threshold 80
334
+ $ fullcourtdefense scan --config .fullcourtdefense.yml --format json
335
+ $ fullcourtdefense credits
336
+ $ fullcourtdefense init
337
+
338
+ \x1b[2mDocs: https://fullcourtdefense.ai/docs/cli\x1b[0m
339
339
  `);
340
340
  }
341
341
  async function main() {
package/dist/version.json CHANGED
@@ -1,3 +1,3 @@
1
1
  {
2
- "version": "1.7.26"
2
+ "version": "1.7.28"
3
3
  }
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "fullcourtdefense-cli",
3
- "version": "1.7.26",
3
+ "version": "1.7.28",
4
4
  "description": "Full Court Defense CLI — security scanning for AI agents from your terminal",
5
5
  "main": "dist/index.js",
6
6
  "bin": {