diff --git a/CLAUDE.md b/CLAUDE.md
index aa7d363..25f417d 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -34,7 +34,7 @@ The CLI is organized as thin layers; command modules call a shared HTTP client a
 - **`api.rs`** — `ApiClient`, a small `reqwest` wrapper with `get`/`get_public`/`post`/`put`/`delete`. There are **two backends**: `ApiClient::new()` (PeerLab API) and `ApiClient::new_saimiris()` (probing API) — pick by which platform the command targets. Authed methods inject the bearer token and validate expiry first (deleting tokens and erroring if expired); `get_public` skips auth.
 - **`auth.rs`** — Auth0 **OAuth2 Device Authorization flow**: `start_device_flow` → `poll_for_token` (handles `authorization_pending`/`slow_down`) → tokens. `refresh_access_token` renews using the stored refresh token. Returns `(access_token, refresh_token, expires_at_unix_secs)`.
 - **`config.rs`** — token persistence to `tokens.json` in the OS config dir (`directories::ProjectDirs::from("dev", "nxthdr", "nxthdr")`). Expiry is enforced client-side from `expires_at`.
-- **`output.rs`** — the single output abstraction. A thread-local `OutputFormat` (set once in `main`) makes every helper format-aware: `section`/`info`/`success`/`warn`/`hint` render only in text mode, while `kv` and `table` emit aligned text **or** structured JSON. Auto-sizes column widths in text mode.
+- **`output.rs`** — the single output abstraction. A thread-local `OutputFormat` (set once in `main`) makes every helper format-aware across three formats (`text`/`json`/`csv`): `section`/`info`/`success`/`warn`/`hint` render only in text mode (gate any text-only decoration or truncation on `output::is_text()`), while `kv` and `table` emit aligned text, structured JSON, **or** RFC-4180 CSV (a `kv` block becomes a single-row CSV; `table` is header + rows). Auto-sizes column widths in text mode.
 - **`peering.rs`** — peering commands against the PeerLab API (`/api/user/info`, `/api/user/prefix`, …), plus `routes`/`lookup` which layer on `ris.rs` for BGP visibility.
 - **`probing.rs`** — probing commands against the Saimiris API, plus `results` which queries ClickHouse directly.
 - **`ris.rs`** — RIPEstat looking-glass client (15s timeout since it's an external public service). Aggregates RIS data into `Visibility` (peer/collector counts, origins, AS paths) and computes propagation % against the full-feed peer count.
@@ -42,7 +42,7 @@ The CLI is organized as thin layers; command modules call a shared HTTP client a
 ## Conventions specific to this repo
 
 - **Request/response types are declared inline inside each command function**, not in a shared models module (e.g. a local `#[derive(Deserialize)] struct UserInfo` scoped to `asn()`). Follow this — only hoist a type when genuinely shared. Each function deserializes just the fields it needs.
-- **Never `println!` user-facing data directly** — route everything through `output.rs` so `-o json` keeps working. The deliberate exceptions are output that is machine-data-by-design: `peering peerlab env` writes a dotenv file to stdout, and empty results print a literal `[]` guarded by `output::is_json()`. Every command must handle its empty/“nothing found” case under both formats.
+- **Never `println!` user-facing data directly** — route everything through `output.rs` so `-o json`/`-o csv` keep working. The deliberate exception is output that is machine-data-by-design: `peering peerlab env` writes a dotenv file to stdout. Every command must handle its empty/“nothing found” case via `output::empty(headers)`, which prints `[]` (JSON) or a header-only row (CSV) and returns `true`, or returns `false` in text mode so the caller can emit a friendly note + hint.
 - **Errors use `anyhow`** end to end (`.context(...)`, `anyhow::bail!`, `anyhow::ensure!`); they bubble up to `main`. Attach user-actionable context (e.g. suggest the `nxthdr` command to run next via `output::hint`).
 - **`tracing::debug!`** is the channel for diagnostic detail (URLs, derived values); surface it with `-v`.
 
diff --git a/src/main.rs b/src/main.rs
index 5cd29d9..46caf3f 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -71,6 +71,11 @@ enum ProbingCommands {
         #[arg(long, help = "Override source IPv6 address (auto-detected per agent if not set)")]
         src_ip: Option<String>,
     },
+    #[command(about = "List your recent measurements")]
+    Measurements {
+        #[arg(long, default_value_t = 20, help = "Maximum number of measurements to list (1-100)")]
+        limit: u32,
+    },
     #[command(about = "Get status of a measurement by ID")]
     MeasurementStatus {
         #[arg(help = "Measurement ID returned by 'send'")]
@@ -170,6 +175,7 @@ async fn handle_probing(command: ProbingCommands) -> anyhow::Result<()> {
         ProbingCommands::Agents => probing::agents().await,
         ProbingCommands::Send { file, agent, src_ip } => probing::send(file, agent, src_ip).await,
         ProbingCommands::Results { src_ip, since, until } => probing::results(src_ip, since, until).await,
+        ProbingCommands::Measurements { limit } => probing::measurements(limit).await,
         ProbingCommands::MeasurementStatus { id } => probing::measurement_status(&id).await,
     }
 }
diff --git a/src/output.rs b/src/output.rs
index 3852ef8..db872a2 100644
--- a/src/output.rs
+++ b/src/output.rs
@@ -6,6 +6,8 @@ pub enum OutputFormat {
     Text,
     #[value(name = "json")]
     Json,
+    #[value(name = "csv")]
+    Csv,
 }
 
 thread_local! {
@@ -16,14 +18,47 @@ pub fn set_format(fmt: OutputFormat) {
     FORMAT.with(|f| f.set(fmt));
 }
 
-pub fn is_json() -> bool {
-    FORMAT.with(|f| f.get()) == OutputFormat::Json
+/// True only in text mode — use to gate human-facing decoration (notes, hints,
+/// truncation) that must not pollute machine-readable JSON/CSV output.
+pub fn is_text() -> bool {
+    FORMAT.with(|f| f.get()) == OutputFormat::Text
 }
 
 fn fmt() -> OutputFormat {
     FORMAT.with(|f| f.get())
 }
 
+/// Escape a single CSV field per RFC 4180: quote it when it contains a comma,
+/// quote, or newline, doubling any embedded quotes.
+fn csv_field(s: &str) -> String {
+    if s.contains([',', '"', '\n', '\r']) {
+        format!("\"{}\"", s.replace('"', "\"\""))
+    } else {
+        s.to_string()
+    }
+}
+
+fn csv_line(cells: &[&str]) -> String {
+    cells.iter().map(|c| csv_field(c)).collect::<Vec<_>>().join(",")
+}
+
+/// Render an empty result set for the active format. In text mode it does
+/// nothing and returns `false` so the caller can print a friendly note; in
+/// JSON it prints `[]` and in CSV the header row, returning `true`.
+pub fn empty(headers: &[&str]) -> bool {
+    match fmt() {
+        OutputFormat::Text => false,
+        OutputFormat::Json => {
+            println!("[]");
+            true
+        }
+        OutputFormat::Csv => {
+            println!("{}", csv_line(headers));
+            true
+        }
+    }
+}
+
 /// Print a section heading with a separator (text mode only).
 pub fn section(title: &str) {
     if fmt() == OutputFormat::Text {
@@ -83,6 +118,14 @@ pub fn kv(pairs: &[(&str, &str)]) {
                 serde_json::to_string_pretty(&serde_json::Value::Object(obj)).unwrap()
             );
         }
+        OutputFormat::Csv => {
+            // A key/value block becomes a single-row CSV: keys as the header,
+            // values as the one data row.
+            let keys: Vec<&str> = pairs.iter().map(|(k, _)| *k).collect();
+            let vals: Vec<&str> = pairs.iter().map(|(_, v)| *v).collect();
+            println!("{}", csv_line(&keys));
+            println!("{}", csv_line(&vals));
+        }
     }
 }
 
@@ -135,5 +178,39 @@ pub fn table(headers: &[&str], rows: &[Vec<String>]) -> usize {
             );
             0
         }
+        OutputFormat::Csv => {
+            println!("{}", csv_line(headers));
+            for row in rows {
+                let cells: Vec<&str> = row.iter().map(|s| s.as_str()).collect();
+                println!("{}", csv_line(&cells));
+            }
+            0
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{csv_field, csv_line};
+
+    #[test]
+    fn csv_field_leaves_plain_values_unquoted() {
+        assert_eq!(csv_field("vltcdg01"), "vltcdg01");
+        assert_eq!(csv_field("2001:db8::/48"), "2001:db8::/48");
+    }
+
+    #[test]
+    fn csv_field_quotes_and_escapes_special_chars() {
+        // Comma forces quoting.
+        assert_eq!(csv_field("a,b"), "\"a,b\"");
+        // Embedded quotes are doubled and the field is wrapped.
+        assert_eq!(csv_field("say \"hi\""), "\"say \"\"hi\"\"\"");
+        // Newlines force quoting.
+        assert_eq!(csv_field("line1\nline2"), "\"line1\nline2\"");
+    }
+
+    #[test]
+    fn csv_line_joins_escaped_fields() {
+        assert_eq!(csv_line(&["a", "b,c", "d"]), "a,\"b,c\",d");
     }
 }
diff --git a/src/peering.rs b/src/peering.rs
index 4a87085..04297bd 100644
--- a/src/peering.rs
+++ b/src/peering.rs
@@ -38,7 +38,7 @@ pub async fn prefix_list() -> anyhow::Result<()> {
     let user_info: UserInfo = api::ApiClient::new().get("/api/user/info").await?;
 
     if user_info.active_leases.is_empty() {
-        if output::is_json() { println!("[]"); } else {
+        if !output::empty(&["prefix", "expires", "rpki"]) {
             output::info("no active prefix leases");
             output::hint("nxthdr peering prefix request <hours>");
         }
@@ -175,9 +175,15 @@ pub async fn routes() -> anyhow::Result<()> {
     let user_info: UserInfo = api::ApiClient::new().get("/api/user/info").await?;
 
     if user_info.active_leases.is_empty() {
-        if output::is_json() {
-            println!("[]");
-        } else {
+        if !output::empty(&[
+            "prefix",
+            "visible",
+            "propagation",
+            "collectors",
+            "peers",
+            "origin",
+            "shortest path",
+        ]) {
             output::info("no active prefix leases — nothing to announce");
             output::hint("nxthdr peering prefix request <hours>");
         }
@@ -215,7 +221,7 @@ pub async fn routes() -> anyhow::Result<()> {
         &rows,
     );
 
-    if !output::is_json() {
+    if output::is_text() {
         let suffix = as_of.map(|t| format!(" (as of {t})")).unwrap_or_default();
         // AS215011 is PeerLab's export ASN; user (private) ASNs are stripped on export.
         output::info(&format!(
@@ -233,15 +239,13 @@ pub async fn lookup(prefix: &str) -> anyhow::Result<()> {
     let vis = ris::looking_glass(prefix).await?;
 
     if !vis.is_visible() {
-        if output::is_json() {
-            println!("[]");
-        } else {
+        if !output::empty(&["origin", "as_path", "peers", "collectors"]) {
             output::info(&format!("{prefix} is not visible in any RIPE RIS collector"));
         }
         return Ok(());
     }
 
-    if !output::is_json() {
+    if output::is_text() {
         let origins = vis.origins();
         let origin_str = if origins.is_empty() { "-".to_string() } else { origins.join(", ") };
         let collectors = vis.collector_count().to_string();
@@ -270,8 +274,8 @@ pub async fn lookup(prefix: &str) -> anyhow::Result<()> {
     }
 
     let paths = vis.paths();
-    // JSON is machine-consumed, so emit every path; the terminal table is capped.
-    let shown = if output::is_json() { paths.len() } else { paths.len().min(20) };
+    // Machine formats (JSON/CSV) emit every path; the terminal table is capped.
+    let shown = if output::is_text() { paths.len().min(20) } else { paths.len() };
     let rows: Vec<Vec<String>> = paths
         .iter()
         .take(shown)
@@ -287,7 +291,7 @@ pub async fn lookup(prefix: &str) -> anyhow::Result<()> {
 
     output::table(&["origin", "as_path", "peers", "collectors"], &rows);
 
-    if !output::is_json() && paths.len() > shown {
+    if output::is_text() && paths.len() > shown {
         output::info(&format!("\n… and {} more distinct paths", paths.len() - shown));
     }
 
diff --git a/src/probing.rs b/src/probing.rs
index 42d5b26..1c00e1a 100644
--- a/src/probing.rs
+++ b/src/probing.rs
@@ -49,7 +49,9 @@ pub async fn agents() -> anyhow::Result<()> {
     let agents: Vec<Agent> = api::ApiClient::new_saimiris().get_public("/api/agents").await?;
 
     if agents.is_empty() {
-        if output::is_json() { println!("[]"); } else { output::info("no agents available"); }
+        if !output::empty(&["id", "status", "prefixes"]) {
+            output::info("no agents available");
+        }
         return Ok(());
     }
 
@@ -226,7 +228,9 @@ pub async fn results(
     let rows = query_clickhouse(&sql).await?;
 
     if rows.is_empty() {
-        if output::is_json() { println!("[]"); } else { output::info(&format!("no replies found for {}", src_ips.join(", "))); }
+        if !output::empty(&["agent", "src", "dst", "ttl", "reply", "rtt"]) {
+            output::info(&format!("no replies found for {}", src_ips.join(", ")));
+        }
         return Ok(());
     }
 
@@ -264,6 +268,49 @@ async fn query_clickhouse(sql: &str) -> anyhow::Result<Vec<serde_json::Value>> {
         .collect()
 }
 
+pub async fn measurements(limit: u32) -> anyhow::Result<()> {
+    anyhow::ensure!((1..=100).contains(&limit), "--limit must be between 1 and 100");
+
+    #[derive(Deserialize)]
+    struct Measurement {
+        measurement_id: String,
+        total_agents: i64,
+        completed_agents: i64,
+        total_expected_probes: i64,
+        total_sent_probes: i64,
+        measurement_complete: bool,
+        started_at: String,
+    }
+
+    let measurements: Vec<Measurement> = api::ApiClient::new_saimiris()
+        .get(&format!("/api/measurements?limit={limit}"))
+        .await?;
+
+    if measurements.is_empty() {
+        if !output::empty(&["id", "started", "agents", "probes", "status"]) {
+            output::info("no measurements found");
+            output::hint("nxthdr probing send --agent <id> probes.csv");
+        }
+        return Ok(());
+    }
+
+    let rows: Vec<Vec<String>> = measurements.iter().map(|m| {
+        let status = if m.measurement_complete { "complete" } else { "in progress" };
+        vec![
+            m.measurement_id.clone(),
+            m.started_at.clone(),
+            format!("{}/{}", m.completed_agents, m.total_agents),
+            format!("{}/{}", m.total_sent_probes, m.total_expected_probes),
+            status.to_string(),
+        ]
+    }).collect();
+
+    output::table(&["id", "started", "agents", "probes", "status"], &rows);
+    output::hint("nxthdr probing measurement-status <id>");
+
+    Ok(())
+}
+
 pub async fn measurement_status(id: &str) -> anyhow::Result<()> {
     #[derive(serde::Deserialize)]
     struct AgentStatus {
@@ -288,14 +335,19 @@ pub async fn measurement_status(id: &str) -> anyhow::Result<()> {
         .get(&format!("/api/measurement/{id}/status"))
         .await?;
 
-    let overall = if status.measurement_complete { "complete" } else { "in progress" };
-    output::section("measurement");
-    output::kv(&[
-        ("id", &status.measurement_id),
-        ("status", overall),
-        ("agents", &format!("{}/{} complete", status.completed_agents, status.total_agents)),
-        ("probes", &format!("{}/{} sent", status.total_sent_probes, status.total_expected_probes)),
-    ]);
+    // Text mode shows the rich summary block; machine formats (json/csv) emit
+    // only the per-agent table below, so the output stays a single valid block
+    // (one CSV header / one JSON value) instead of a summary followed by a table.
+    if output::is_text() {
+        let overall = if status.measurement_complete { "complete" } else { "in progress" };
+        output::section("measurement");
+        output::kv(&[
+            ("id", &status.measurement_id),
+            ("status", overall),
+            ("agents", &format!("{}/{} complete", status.completed_agents, status.total_agents)),
+            ("probes", &format!("{}/{} sent", status.total_sent_probes, status.total_expected_probes)),
+        ]);
+    }
 
     if !status.agents.is_empty() {
         let rows: Vec<Vec<String>> = status.agents.iter().map(|a| {