Add upstream proxy

Cargo.toml

@@ -27,15 +27,18 @@ webpki-roots = "0.26"
 lru = "0.12"
 rand = "0.8"
 anyhow = "1.0"
+base64 = "0.22"
 tracing = "0.1"
 tracing-subscriber = { version = "0.3", features = ["env-filter", "chrono"] }
 chrono = "0.4"
 dirs = "6.0.0"
 hyper-rustls = "0.27.7"
+tower-service = "0.3"
 tls-parser = "0.12.2"
 camino = "1.1.11"
 filetime = "0.2"
 ctrlc = "3.4"
+percent-encoding = "2.3"
 url = "2.5"
 v8 = "129"
 serde = { version = "1.0", features = ["derive"] }

README.md

@@ -23,6 +23,7 @@ Or download a pre-built binary from the [releases page](https://github.com/coder
 - 🌐 **HTTP/HTTPS interception** - Transparent proxy with TLS certificate injection
 - 🛡️ **DNS exfiltration protection** - Prevents data leakage through DNS queries
 - 🔧 **Multiple evaluation approaches** - JS expressions or custom programs
+- 🏢 **Upstream proxy support** - Chain httpjail's egress through a corporate proxy
 - 🖥️ **Cross-platform** - Native support for Linux and macOS
 
 ## Quick Start
@@ -61,8 +62,28 @@ httpjail --server --js "true"
 
 # Run Docker containers with network isolation (Linux only)
 httpjail --js "r.host === 'api.github.com'" --docker-run -- --rm alpine:latest wget -qO- https://api.github.com
+
+# Route httpjail's own egress through an upstream (corporate) proxy
+httpjail --upstream-proxy http://proxy.corp:3128 --js "true" -- curl https://api.github.com
+# Credentials and HTTPS proxies are supported: http://user:pass@proxy.corp:3128, https://proxy.corp:8443
+# May also be set via the HTTPJAIL_UPSTREAM_PROXY environment variable
 ```
 
+### Upstream (corporate) proxy
+
+When httpjail itself runs in an environment with no direct internet access, use
+`--upstream-proxy <URL>` (or the `HTTPJAIL_UPSTREAM_PROXY` environment variable)
+to route httpjail's outbound requests through an upstream proxy. Rule evaluation
+still happens locally on the intercepted traffic; only the re-originated request
+is forwarded through the proxy.
+
+- `http://`, `https://` and bare `host:port` (http assumed) forms are accepted.
+- Basic authentication is supported via `http://user:pass@host:port`.
+- HTTPS destinations are reached via a `CONNECT` tunnel through the proxy, while
+  plain HTTP destinations are forwarded in absolute-form.
+- This is independent of the `HTTP_PROXY`/`HTTPS_PROXY` variables that httpjail
+  sets *inside* the jail to point sandboxed processes at itself.
+
 ## Documentation
 
 Docs are stored in the `docs/` directory and served
@@ -82,6 +103,7 @@ Table of Contents:
 - [TLS Interception](https://coder.github.io/httpjail/advanced/tls-interception.html)
 - [DNS Exfiltration](https://coder.github.io/httpjail/advanced/dns-exfiltration.html)
 - [Server Mode](https://coder.github.io/httpjail/advanced/server-mode.html)
+- [Upstream Proxy](https://coder.github.io/httpjail/advanced/upstream-proxy.html)
 
 ## License
 

docs/SUMMARY.md

@@ -20,6 +20,7 @@
 - [TLS Interception](./advanced/tls-interception.md)
 - [DNS Exfiltration](./advanced/dns-exfiltration.md)
 - [Server Mode](./advanced/server-mode.md)
+- [Upstream Proxy](./advanced/upstream-proxy.md)
 
 ---
 

docs/advanced/upstream-proxy.md

@@ -0,0 +1,61 @@
+# Upstream Proxy
+
+By default httpjail contacts destination servers directly. When httpjail itself
+runs in an environment that has no direct internet access — for example behind a
+corporate proxy — you can route httpjail's own outbound requests through an
+upstream proxy with `--upstream-proxy` (or the `HTTPJAIL_UPSTREAM_PROXY`
+environment variable).
+
+Rule evaluation still happens locally on the intercepted traffic. Only the
+request that httpjail re-originates towards the real destination is forwarded
+through the upstream proxy.
+
+```bash
+# Route httpjail's egress through a corporate proxy
+httpjail --upstream-proxy http://proxy.corp:3128 --js "true" -- curl https://api.github.com
+
+# With Basic authentication
+httpjail --upstream-proxy http://user:pass@proxy.corp:3128 --js "true" -- ./my-app
+
+# Through an HTTPS proxy
+httpjail --upstream-proxy https://proxy.corp:8443 --js "true" -- ./my-app
+
+# Via the environment variable (equivalent to --upstream-proxy)
+HTTPJAIL_UPSTREAM_PROXY=http://proxy.corp:3128 httpjail --js "true" -- ./my-app
+```
+
+## Accepted formats
+
+| Form | Example | Notes |
+| --- | --- | --- |
+| `http://host:port` | `http://proxy.corp:3128` | Plain HTTP proxy |
+| `https://host:port` | `https://proxy.corp:8443` | Connection to the proxy is wrapped in TLS |
+| `host:port` | `proxy.corp:3128` | Bare authority, `http` scheme assumed |
+| With credentials | `http://user:pass@proxy.corp:3128` | Sends `Proxy-Authorization: Basic ...` |
+
+The command-line flag takes precedence over the environment variable. Credentials
+are never written to the logs.
+
+## How it works
+
+- **HTTPS destinations** are reached by issuing a `CONNECT` to the upstream
+  proxy to obtain a raw TCP tunnel; httpjail then performs the destination TLS
+  handshake over that tunnel. TLS is validated against Mozilla's webpki roots
+  plus the httpjail CA, exactly as for a direct connection.
+- **Plain HTTP destinations** are forwarded to the proxy in absolute-form, with
+  the `Proxy-Authorization` header attached when credentials are configured.
+- Only connection setup (TCP connect, optional TLS to the proxy, and the
+  `CONNECT` exchange) is bounded by a timeout. The established tunnel carries no
+  timeout, so long-running connections such as WebSocket and gRPC keep working.
+
+## Relationship to `HTTP_PROXY` / `HTTPS_PROXY`
+
+This feature is independent of the `HTTP_PROXY` and `HTTPS_PROXY` variables that
+httpjail sets *inside* the jail to point sandboxed processes at httpjail itself.
+
+```
+[ jailed process ] --HTTP_PROXY/HTTPS_PROXY--> [ httpjail ] --upstream-proxy--> [ corporate proxy ] --> internet
+```
+
+The jailed process always talks to httpjail; `--upstream-proxy` only affects the
+hop from httpjail to the outside world.

docs/guide/configuration.md

@@ -88,10 +88,15 @@ These are automatically set in the jailed process:
 
 These affect httpjail's behavior:
 
-| Variable           | Description                | Example                          |
-| ------------------ | -------------------------- | -------------------------------- |
-| `RUST_LOG`         | Logging level              | `debug`, `info`, `warn`, `error` |
-| `HTTPJAIL_CA_CERT` | Custom CA certificate path | `/etc/pki/custom-ca.pem`         |
+| Variable                 | Description                            | Example                          |
+| ------------------------ | -------------------------------------- | -------------------------------- |
+| `RUST_LOG`               | Logging level                          | `debug`, `info`, `warn`, `error` |
+| `HTTPJAIL_CA_CERT`       | Custom CA certificate path             | `/etc/pki/custom-ca.pem`         |
+| `HTTPJAIL_UPSTREAM_PROXY`| Upstream proxy for httpjail's egress   | `http://proxy.corp:3128`         |
+
+The `--upstream-proxy` command-line flag takes precedence over
+`HTTPJAIL_UPSTREAM_PROXY`. See [Upstream Proxy](../advanced/upstream-proxy.md)
+for details.
 
 ## Platform-Specific Configuration
 

src/lib.rs

@@ -9,5 +9,6 @@ pub mod proxy_tls;
 pub mod rules;
 pub mod sys_resource;
 pub mod tls;
+pub mod upstream;
 
 pub mod test_utils;

src/main.rs

@@ -7,6 +7,7 @@ use httpjail::rules::shell::ShellRuleEngine;
 use httpjail::rules::v8_js::V8JsRuleEngine;
 use httpjail::rules::{Action, RuleEngine};
 use hyper::Method;
+use std::fmt;
 use std::fs::OpenOptions;
 use std::os::unix::process::ExitStatusExt;
 use std::sync::atomic::{AtomicBool, Ordering};
@@ -40,7 +41,7 @@ enum Command {
     },
 }
 
-#[derive(Parser, Debug)]
+#[derive(Parser)]
 struct RunArgs {
     /// Use shell script for evaluating requests
     /// The script receives environment variables:
@@ -85,6 +86,13 @@ struct RunArgs {
     #[arg(long = "request-log", value_name = "FILE")]
     request_log: Option<String>,
 
+    /// Route httpjail's own upstream requests through an upstream (corporate) proxy.
+    /// Accepts http://host:port, https://host:port, or host:port (http assumed),
+    /// optionally with credentials: http://user:pass@host:port.
+    /// Falls back to the HTTPJAIL_UPSTREAM_PROXY environment variable.
+    #[arg(long = "upstream-proxy", value_name = "URL")]
+    upstream_proxy: Option<String>,
+
     /// Use weak mode (environment variables only, no system isolation)
     #[arg(long = "weak")]
     weak: bool,
@@ -139,6 +147,32 @@ struct RunArgs {
     exec_command: Vec<String>,
 }
 
+impl fmt::Debug for RunArgs {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        let upstream_proxy = self
+            .upstream_proxy
+            .as_deref()
+            .map(httpjail::upstream::redact_proxy_spec);
+        f.debug_struct("RunArgs")
+            .field("sh", &self.sh)
+            .field("proc", &self.proc)
+            .field("js", &self.js)
+            .field("js_file", &self.js_file)
+            .field("request_log", &self.request_log)
+            .field("upstream_proxy", &upstream_proxy)
+            .field("weak", &self.weak)
+            .field("verbose", &self.verbose)
+            .field("timeout", &self.timeout)
+            .field("no_jail_cleanup", &self.no_jail_cleanup)
+            .field("cleanup", &self.cleanup)
+            .field("server", &self.server)
+            .field("test", &self.test)
+            .field("docker_run", &self.docker_run)
+            .field("exec_command", &self.exec_command)
+            .finish()
+    }
+}
+
 fn setup_logging(verbosity: u8) {
     use tracing_subscriber::fmt::time::FormatTime;
 
@@ -590,7 +624,27 @@ async fn main() -> Result<()> {
         }
     };
 
-    let mut proxy = ProxyServer::new(http_bind, https_bind, rule_engine);
+    // Resolve the optional upstream (corporate) proxy from the flag or env var.
+    // This is independent of the HTTP_PROXY/HTTPS_PROXY variables httpjail sets
+    // *inside* the jail to point sandboxed processes at itself.
+    let upstream_proxy_spec = args
+        .run_args
+        .upstream_proxy
+        .clone()
+        .or_else(|| std::env::var("HTTPJAIL_UPSTREAM_PROXY").ok());
+    let upstream_proxy = match upstream_proxy_spec {
+        Some(spec) => {
+            let proxy = httpjail::upstream::UpstreamProxy::parse(&spec)
+                .with_context(|| format!("Failed to parse upstream proxy: {}", spec))?;
+            // Avoid logging the spec verbatim as it may contain credentials.
+            info!("Routing httpjail upstream requests through the configured upstream proxy");
+            Some(proxy)
+        }
+        None => None,
+    };
+
+    let mut proxy =
+        ProxyServer::new_with_upstream_proxy(http_bind, https_bind, rule_engine, upstream_proxy);
 
     // Start proxy in background if running as server; otherwise start with random ports
     let (actual_http_port, actual_https_port) = proxy.start().await?;