Skip to main content

browser_control/mcp/
tools.rs

1//! MCP tools exposed by the `browser-control mcp` server.
2//!
3//! Ten tools wrap the underlying `session::PageSession` and related helpers:
4//! `navigate`, `get_dom`, `screenshot`, `fetch`, `select_element`,
5//! `list_targets`, `cookies`, `storage_get`, `storage_set`, `wait_for_cookie`.
6
7use anyhow::{anyhow, Result};
8use regex::Regex;
9use serde_json::{json, Value};
10use std::sync::Arc;
11use std::time::{Duration, Instant};
12
13use crate::cli::cookies::fetch_cookies;
14use crate::cli::storage::{build_get_expr, build_set_expr, ns_global};
15use crate::cli::wait_for_cookie::cookie_matches;
16use crate::detect::Engine;
17use crate::dom::scripts::{FETCH_JS, GET_DOM_JS, SELECT_ELEMENT_JS};
18use crate::mcp::server::{RegisteredTool, ServerState, ToolHandler, ToolRegistry};
19use crate::session::attach::PageSession;
20use crate::session::targets::{list as list_targets, open_bidi};
21
22/// Register the standard tool set onto the given registry.
23pub fn register_all(registry: &ToolRegistry) {
24    registry.register(make_navigate());
25    registry.register(make_get_dom());
26    registry.register(make_screenshot());
27    registry.register(make_fetch());
28    registry.register(make_select_element());
29    registry.register(make_list_targets());
30    registry.register(make_cookies());
31    registry.register(make_storage_get());
32    registry.register(make_storage_set());
33    registry.register(make_wait_for_cookie());
34}
35
36// ---------------------------------------------------------------------------
37// Engine dispatch helpers.
38// ---------------------------------------------------------------------------
39
40/// Open (or return the cached) page session for the configured browser.
41///
42/// CDP sessions are short-lived: we open and close a fresh WebSocket per
43/// call. BiDi sessions reuse one persistent client, because Firefox limits
44/// a browser instance to a single concurrent BiDi session.
45async fn attach_active(state: &ServerState) -> Result<PageSession> {
46    match state.browser.engine {
47        Engine::Cdp => PageSession::attach(&state.browser.endpoint, Engine::Cdp, None).await,
48        Engine::Bidi => {
49            let mut guard = state.bidi.lock().await;
50            let client = if let Some(c) = guard.as_ref() {
51                c.clone()
52            } else {
53                let c = Arc::new(open_bidi(&state.browser.endpoint).await?);
54                c.session_new().await?;
55                *guard = Some(c.clone());
56                c
57            };
58            PageSession::from_bidi_cache(client, None).await
59        }
60    }
61}
62
63fn text_content(text: impl Into<String>) -> Value {
64    json!({ "content": [ { "type": "text", "text": text.into() } ] })
65}
66
67fn image_content(data: String) -> Value {
68    json!({
69        "content": [ { "type": "image", "data": data, "mimeType": "image/png" } ]
70    })
71}
72
73fn handler<F>(f: F) -> ToolHandler
74where
75    F: Fn(ServerState, Value) -> futures_util::future::BoxFuture<'static, Result<Value>>
76        + Send
77        + Sync
78        + 'static,
79{
80    Arc::new(f)
81}
82
83// ---------------------------------------------------------------------------
84// navigate
85// ---------------------------------------------------------------------------
86
87fn make_navigate() -> RegisteredTool {
88    RegisteredTool {
89        name: "navigate".into(),
90        description: "Navigate the active page to a URL.".into(),
91        input_schema: json!({
92            "type": "object",
93            "properties": { "url": { "type": "string" } },
94            "required": ["url"],
95        }),
96        handler: handler(|state, args| {
97            Box::pin(async move {
98                let url = args
99                    .get("url")
100                    .and_then(|v| v.as_str())
101                    .ok_or_else(|| anyhow!("missing 'url'"))?
102                    .to_string();
103                let session = attach_active(&state).await?;
104                session.navigate(&url).await?;
105                session.close().await;
106                Ok(text_content(format!("Navigated to {url}")))
107            })
108        }),
109    }
110}
111
112// ---------------------------------------------------------------------------
113// get_dom
114// ---------------------------------------------------------------------------
115
116fn make_get_dom() -> RegisteredTool {
117    RegisteredTool {
118        name: "get_dom".into(),
119        description: "Get the rendered DOM as HTML, with shadow roots serialized when supported."
120            .into(),
121        input_schema: json!({
122            "type": "object",
123            "properties": {
124                "selector": {
125                    "type": "string",
126                    "description": "Optional CSS selector; defaults to the document element."
127                }
128            },
129        }),
130        handler: handler(|state, args| {
131            Box::pin(async move {
132                let selector_arg = args.get("selector").and_then(|v| v.as_str());
133                let selector_literal = match selector_arg {
134                    Some(s) => serde_json::to_string(s)?,
135                    None => "null".to_string(),
136                };
137                let expr = format!("({GET_DOM_JS})({selector_literal})");
138                let session = attach_active(&state).await?;
139                let value = session.evaluate(&expr, false).await?;
140                session.close().await;
141                let html = value.as_str().unwrap_or("").to_string();
142                Ok(text_content(html))
143            })
144        }),
145    }
146}
147
148// ---------------------------------------------------------------------------
149// screenshot
150// ---------------------------------------------------------------------------
151
152fn make_screenshot() -> RegisteredTool {
153    RegisteredTool {
154        name: "screenshot".into(),
155        description: "Capture a PNG screenshot of the active page.".into(),
156        input_schema: json!({
157            "type": "object",
158            "properties": {
159                "full_page": { "type": "boolean", "default": false },
160                "selector": { "type": "string" }
161            },
162        }),
163        handler: handler(|state, args| {
164            Box::pin(async move {
165                let full_page = args
166                    .get("full_page")
167                    .and_then(|v| v.as_bool())
168                    .unwrap_or(false);
169                let session = attach_active(&state).await?;
170                let b64 = session.screenshot(full_page).await?;
171                session.close().await;
172                Ok(image_content(b64))
173            })
174        }),
175    }
176}
177
178// ---------------------------------------------------------------------------
179// fetch
180// ---------------------------------------------------------------------------
181
182fn make_fetch() -> RegisteredTool {
183    RegisteredTool {
184        name: "fetch".into(),
185        description:
186            "Perform an HTTP request from the page context (preserves cookies, bypasses CORS)."
187                .into(),
188        input_schema: json!({
189            "type": "object",
190            "properties": {
191                "url": { "type": "string" },
192                "method": { "type": "string" },
193                "headers": { "type": "object" },
194                "body": { "type": "string" }
195            },
196            "required": ["url"],
197        }),
198        handler: handler(|state, args| {
199            Box::pin(async move {
200                if args.get("url").and_then(|v| v.as_str()).is_none() {
201                    return Err(anyhow!("missing 'url'"));
202                }
203                let args_json = serde_json::to_string(&args)?;
204                let args_literal = serde_json::to_string(&args_json)?;
205                let expr = format!("({FETCH_JS})({args_literal})");
206                let session = attach_active(&state).await?;
207                let value = session.evaluate(&expr, true).await?;
208                session.close().await;
209                let raw = value.as_str().unwrap_or("").to_string();
210                let parsed: Value = serde_json::from_str(&raw)
211                    .map_err(|e| anyhow!("invalid fetch response JSON: {e}"))?;
212                let pretty = serde_json::to_string_pretty(&parsed)?;
213                Ok(text_content(pretty))
214            })
215        }),
216    }
217}
218
219// ---------------------------------------------------------------------------
220// select_element
221// ---------------------------------------------------------------------------
222
223fn make_select_element() -> RegisteredTool {
224    RegisteredTool {
225        name: "select_element".into(),
226        description:
227            "Show an interactive overlay; resolve with the CSS selector for the clicked element."
228                .into(),
229        input_schema: json!({
230            "type": "object",
231            "properties": {},
232        }),
233        handler: handler(|state, _args| {
234            Box::pin(async move {
235                let expr = SELECT_ELEMENT_JS.to_string();
236                let session = attach_active(&state).await?;
237                let value = session.evaluate(&expr, true).await?;
238                session.close().await;
239                let selector = value.as_str().unwrap_or("").to_string();
240                Ok(text_content(selector))
241            })
242        }),
243    }
244}
245
246// ---------------------------------------------------------------------------
247// list_targets
248// ---------------------------------------------------------------------------
249
250fn make_list_targets() -> RegisteredTool {
251    RegisteredTool {
252        name: "list_targets".into(),
253        description: "List open page targets, optionally filtered by an unanchored URL regex."
254            .into(),
255        input_schema: json!({
256            "type": "object",
257            "properties": {
258                "filter": {
259                    "type": "string",
260                    "description": "Optional unanchored URL regex."
261                }
262            },
263        }),
264        handler: handler(|state, args| {
265            Box::pin(async move {
266                let filter = args
267                    .get("filter")
268                    .and_then(|v| v.as_str())
269                    .map(|s| s.to_string());
270                let targets =
271                    list_targets(&state.browser.endpoint, state.browser.engine, filter.as_deref())
272                        .await?;
273                Ok(text_content(serde_json::to_string_pretty(&targets)?))
274            })
275        }),
276    }
277}
278
279// ---------------------------------------------------------------------------
280// cookies
281// ---------------------------------------------------------------------------
282
283fn make_cookies() -> RegisteredTool {
284    RegisteredTool {
285        name: "cookies".into(),
286        description: "Fetch cookies from the active browser. Returns full values (MCP is a \
287                      trusted local channel). Optional unanchored regex filters."
288            .into(),
289        input_schema: json!({
290            "type": "object",
291            "properties": {
292                "domain": { "type": "string", "description": "Unanchored regex on cookie domain." },
293                "name":   { "type": "string", "description": "Unanchored regex on cookie name." }
294            },
295        }),
296        handler: handler(|state, args| {
297            Box::pin(async move {
298                let domain_re = args
299                    .get("domain")
300                    .and_then(|v| v.as_str())
301                    .map(Regex::new)
302                    .transpose()
303                    .map_err(|e| anyhow!("invalid `domain` regex: {e}"))?;
304                let name_re = args
305                    .get("name")
306                    .and_then(|v| v.as_str())
307                    .map(Regex::new)
308                    .transpose()
309                    .map_err(|e| anyhow!("invalid `name` regex: {e}"))?;
310                let all = fetch_cookies(&state.browser).await?;
311                let filtered: Vec<_> = all
312                    .into_iter()
313                    .filter(|c| {
314                        domain_re.as_ref().map_or(true, |re| re.is_match(&c.domain))
315                            && name_re.as_ref().map_or(true, |re| re.is_match(&c.name))
316                    })
317                    .collect();
318                Ok(text_content(serde_json::to_string_pretty(&filtered)?))
319            })
320        }),
321    }
322}
323
324// ---------------------------------------------------------------------------
325// storage_get / storage_set
326// ---------------------------------------------------------------------------
327
328fn make_storage_get() -> RegisteredTool {
329    RegisteredTool {
330        name: "storage_get".into(),
331        description: "Read a value from localStorage or sessionStorage on the active page.".into(),
332        input_schema: json!({
333            "type": "object",
334            "properties": {
335                "key": { "type": "string" },
336                "namespace": {
337                    "type": "string",
338                    "enum": ["local", "session"],
339                    "default": "local"
340                }
341            },
342            "required": ["key"],
343        }),
344        handler: handler(|state, args| {
345            Box::pin(async move {
346                let key = args
347                    .get("key")
348                    .and_then(|v| v.as_str())
349                    .ok_or_else(|| anyhow!("missing 'key'"))?
350                    .to_string();
351                let namespace = args
352                    .get("namespace")
353                    .and_then(|v| v.as_str())
354                    .unwrap_or("local");
355                let ns = ns_global(namespace)?;
356                let expr = build_get_expr(ns, &key);
357                let session = attach_active(&state).await?;
358                let value = session.evaluate(&expr, true).await?;
359                session.close().await;
360                // `build_get_expr` wraps the result in JSON.stringify, so the
361                // evaluator returns a JSON string. Unwrap one layer to surface
362                // the raw value (or `null` when the key is absent).
363                let text = match value {
364                    Value::String(s) => s,
365                    Value::Null => "null".to_string(),
366                    other => other.to_string(),
367                };
368                Ok(text_content(text))
369            })
370        }),
371    }
372}
373
374fn make_storage_set() -> RegisteredTool {
375    RegisteredTool {
376        name: "storage_set".into(),
377        description: "Write a value to localStorage or sessionStorage on the active page.".into(),
378        input_schema: json!({
379            "type": "object",
380            "properties": {
381                "key": { "type": "string" },
382                "value": { "type": "string" },
383                "namespace": {
384                    "type": "string",
385                    "enum": ["local", "session"],
386                    "default": "local"
387                }
388            },
389            "required": ["key", "value"],
390        }),
391        handler: handler(|state, args| {
392            Box::pin(async move {
393                let key = args
394                    .get("key")
395                    .and_then(|v| v.as_str())
396                    .ok_or_else(|| anyhow!("missing 'key'"))?
397                    .to_string();
398                let value = args
399                    .get("value")
400                    .and_then(|v| v.as_str())
401                    .ok_or_else(|| anyhow!("missing 'value'"))?
402                    .to_string();
403                let namespace = args
404                    .get("namespace")
405                    .and_then(|v| v.as_str())
406                    .unwrap_or("local");
407                let ns = ns_global(namespace)?;
408                let expr = build_set_expr(ns, &key, &value);
409                let session = attach_active(&state).await?;
410                let _ = session.evaluate(&expr, true).await?;
411                session.close().await;
412                Ok(text_content("ok"))
413            })
414        }),
415    }
416}
417
418// ---------------------------------------------------------------------------
419// wait_for_cookie
420// ---------------------------------------------------------------------------
421
422fn make_wait_for_cookie() -> RegisteredTool {
423    RegisteredTool {
424        name: "wait_for_cookie".into(),
425        description: "Poll the browser until a cookie matching the regex filters appears, or \
426                      timeout elapses."
427            .into(),
428        input_schema: json!({
429            "type": "object",
430            "properties": {
431                "domain": { "type": "string", "description": "Unanchored regex on cookie domain." },
432                "name":   { "type": "string", "description": "Unanchored regex on cookie name." },
433                "timeout_seconds": { "type": "number", "default": 120 },
434                "poll_interval_seconds": { "type": "number", "default": 1 }
435            },
436            "required": ["domain", "name"],
437        }),
438        handler: handler(|state, args| {
439            Box::pin(async move {
440                let domain = args
441                    .get("domain")
442                    .and_then(|v| v.as_str())
443                    .ok_or_else(|| anyhow!("missing 'domain'"))?;
444                let name = args
445                    .get("name")
446                    .and_then(|v| v.as_str())
447                    .ok_or_else(|| anyhow!("missing 'name'"))?;
448                let domain_re =
449                    Regex::new(domain).map_err(|e| anyhow!("invalid `domain` regex: {e}"))?;
450                let name_re =
451                    Regex::new(name).map_err(|e| anyhow!("invalid `name` regex: {e}"))?;
452                let timeout_s = args
453                    .get("timeout_seconds")
454                    .and_then(|v| v.as_f64())
455                    .unwrap_or(120.0)
456                    .max(0.0);
457                let interval_s = args
458                    .get("poll_interval_seconds")
459                    .and_then(|v| v.as_f64())
460                    .unwrap_or(1.0)
461                    .max(0.001);
462                let deadline = Instant::now() + Duration::from_secs_f64(timeout_s);
463                let interval = Duration::from_secs_f64(interval_s);
464                loop {
465                    let cookies = fetch_cookies(&state.browser).await?;
466                    if let Some(c) = cookies
467                        .into_iter()
468                        .find(|c| cookie_matches(c, &domain_re, &name_re))
469                    {
470                        return Ok(text_content(c.name));
471                    }
472                    let now = Instant::now();
473                    if now >= deadline {
474                        return Err(anyhow!("timed out waiting for cookie"));
475                    }
476                    let remaining = deadline.saturating_duration_since(now);
477                    let nap = std::cmp::min(interval, remaining);
478                    if nap.is_zero() {
479                        return Err(anyhow!("timed out waiting for cookie"));
480                    }
481                    tokio::time::sleep(nap).await;
482                }
483            })
484        }),
485    }
486}
487
488#[cfg(test)]
489mod tests {
490    use super::*;
491
492    const EXPECTED_TOOLS: &[&str] = &[
493        "navigate",
494        "get_dom",
495        "screenshot",
496        "fetch",
497        "select_element",
498        "list_targets",
499        "cookies",
500        "storage_get",
501        "storage_set",
502        "wait_for_cookie",
503    ];
504
505    fn schema_for(name: &str) -> Value {
506        let registry = ToolRegistry::new();
507        register_all(&registry);
508        registry
509            .list()
510            .into_iter()
511            .find(|t| t["name"] == name)
512            .unwrap_or_else(|| panic!("tool {name} not registered"))["inputSchema"]
513            .clone()
514    }
515
516    #[test]
517    fn register_all_adds_ten_tools() {
518        let registry = ToolRegistry::new();
519        register_all(&registry);
520        let list = registry.list();
521        assert_eq!(list.len(), 10);
522        let names: Vec<&str> = list.iter().map(|t| t["name"].as_str().unwrap()).collect();
523        for expected in EXPECTED_TOOLS {
524            assert!(
525                names.contains(expected),
526                "missing tool {expected} in {names:?}"
527            );
528        }
529    }
530
531    #[test]
532    fn every_tool_has_object_input_schema() {
533        let registry = ToolRegistry::new();
534        register_all(&registry);
535        for t in registry.list() {
536            let schema = &t["inputSchema"];
537            assert!(schema.is_object(), "schema not object: {schema}");
538            assert_eq!(
539                schema["type"], "object",
540                "schema type != object for {}: {schema}",
541                t["name"]
542            );
543        }
544    }
545
546    #[test]
547    fn list_targets_schema_has_optional_filter() {
548        let schema = schema_for("list_targets");
549        assert_eq!(schema["properties"]["filter"]["type"], "string");
550        assert!(schema.get("required").is_none() || schema["required"].as_array().unwrap().is_empty());
551    }
552
553    #[test]
554    fn cookies_schema_has_optional_filters() {
555        let schema = schema_for("cookies");
556        assert_eq!(schema["properties"]["domain"]["type"], "string");
557        assert_eq!(schema["properties"]["name"]["type"], "string");
558        assert!(schema.get("required").is_none() || schema["required"].as_array().unwrap().is_empty());
559    }
560
561    #[test]
562    fn storage_get_requires_key() {
563        let schema = schema_for("storage_get");
564        let required = schema["required"].as_array().expect("required array");
565        assert!(required.iter().any(|v| v == "key"));
566        assert_eq!(schema["properties"]["key"]["type"], "string");
567        assert_eq!(schema["properties"]["namespace"]["type"], "string");
568    }
569
570    #[test]
571    fn storage_set_requires_key_and_value() {
572        let schema = schema_for("storage_set");
573        let required: Vec<&str> = schema["required"]
574            .as_array()
575            .unwrap()
576            .iter()
577            .map(|v| v.as_str().unwrap())
578            .collect();
579        assert!(required.contains(&"key"));
580        assert!(required.contains(&"value"));
581        assert_eq!(schema["properties"]["value"]["type"], "string");
582    }
583
584    #[test]
585    fn wait_for_cookie_requires_domain_and_name() {
586        let schema = schema_for("wait_for_cookie");
587        let required: Vec<&str> = schema["required"]
588            .as_array()
589            .unwrap()
590            .iter()
591            .map(|v| v.as_str().unwrap())
592            .collect();
593        assert!(required.contains(&"domain"));
594        assert!(required.contains(&"name"));
595        assert_eq!(schema["properties"]["timeout_seconds"]["type"], "number");
596        assert_eq!(
597            schema["properties"]["poll_interval_seconds"]["type"],
598            "number"
599        );
600    }
601}