feat: Talk to agents over A2A (agent card / send / task)

README.md

@@ -339,6 +339,54 @@ field works without upgrading the CLI. `--from-version latest|N` starts from an
 existing version and applies your overrides on top, replacing whole top-level
 keys rather than deep-merging.
 
+### Talking to an agent (A2A)
+
+Every agent bound to a workspace answers over the
+[A2A protocol](https://a2a-protocol.org). The CLI speaks A2A v1.0 over
+HTTP+JSON; `--workspace` defaults to the remembered one, as for `agent bind`.
+
+```bash
+memorylake agent card <agent-id> [--workspace <id>]
+
+memorylake agent send <agent-id> --text "..." [--text "..."] \
+  [--context CTX] [--task TASK] [--stream | --no-wait] [--raw] \
+  [--actor ID] [--project ID] [--read-only-project ID]... [--skip-memory] \
+  [--metadata-json '{"overrides":{...}}'] [--workspace <id>]
+memorylake agent send <agent-id> --message-json '[{"text":"..."},{"url":"...","mediaType":"image/png"}]'
+memorylake agent send <agent-id> --message-file parts.json
+
+memorylake agent task list <agent-id> [--context CTX] [--status STATE] \
+  [--page-size N] [--page-token TOK] [--after TIMESTAMP] [--history-length N] [--artifacts]
+memorylake agent task get      <agent-id> <task-id> [--history-length N]
+memorylake agent task cancel   <agent-id> <task-id>
+memorylake agent task feedback <agent-id> <task-id> --rating up|down [--comment TEXT]
+```
+
+`send` waits for the answer and prints the reply text on stdout; the task id,
+context id and final state go to stderr, so a script can pipe the reply and
+still know how to continue:
+
+```
+$ memorylake agent send agent-… --text "Summarize yesterday's standup"
+The team agreed to …
+task run-…  context 5fdb…  state TASK_STATE_COMPLETED
+```
+
+Pass `--context` to keep talking in the same thread. A task that ends in
+`TASK_STATE_INPUT_REQUIRED` needs more from you: reply with `--task <id>
+--context <id>`. `--stream` prints the reply as it is produced; `--no-wait`
+returns the task as soon as it exists (poll it with `agent task get`); `--raw`
+prints the protocol response as JSON instead of the reply text.
+
+The `--actor`, `--project`, `--read-only-project` and `--skip-memory` flags set
+MemoryLake's extension of the request (`metadata.memorylake`): whose message it
+is, which projects the agent may read and write, and whether the exchange is
+remembered at all. Anything else the extension accepts, such as `overrides`,
+goes through `--metadata-json`.
+
+Feedback is a MemoryLake extension to A2A. `task feedback` records a rating on
+the task; `task get` reads it back under `metadata."task-feedback/v1"`.
+
 ### Search
 
 ```bash

crates/cli/src/commands/agent.rs

@@ -2,8 +2,10 @@
 //!
 //! Agent *identity* (name, description, metadata) changes in place via
 //! `agent update`. Agent *configuration* (model, policies, prompt, …) is
-//! immutable and changes only by creating a new version.
+//! immutable and changes only by creating a new version. Talking to a bound
+//! agent (`card`, `send`, `task`) goes over A2A and lives in [`a2a`].
 
+mod a2a;
 mod body;
 
 use anyhow::{Context, Result};
@@ -17,6 +19,7 @@ use memorylake_core::{Client, Paths, ResolveOverrides, resolve};
 use std::path::PathBuf;
 
 use super::require_workspace;
+use a2a::{SendArgs, TaskCommand, run_card, run_send, run_task, task_workspace_flag};
 use body::{FromVersion, load_config_body, reject_config_fields, require_field, set_scalar};
 
 /// Agent subcommands.
@@ -138,6 +141,26 @@ pub enum AgentCommand {
         #[arg(long = "name")]
         name_fuzzy: Option<String>,
     },
+    /// Show an agent's A2A card: capabilities, protocol bindings, extensions.
+    Card {
+        /// Agent id (must be bound to the workspace).
+        agent_id: String,
+        /// Workspace the agent is bound in.
+        ///
+        /// Defaults to the workspace remembered by `workspace use`.
+        #[arg(long)]
+        workspace: Option<String>,
+    },
+    /// Send a message to an agent and print its reply.
+    ///
+    /// Waits for the answer by default. The reply text goes to stdout; the
+    /// task and context ids needed to continue go to stderr.
+    Send(SendArgs),
+    /// Inspect, cancel and rate the tasks an agent has run.
+    Task {
+        #[command(subcommand)]
+        command: TaskCommand,
+    },
 }
 
 /// `agent version` subcommands.
@@ -303,6 +326,22 @@ pub fn run(command: AgentCommand, profile: Option<String>, base_url: Option<Stri
             .with_context(|| format!("list agents bound to workspace `{workspace}`"))?;
             println!("{}", serde_json::to_string_pretty(&data)?);
         }
+        AgentCommand::Card {
+            agent_id,
+            workspace,
+        } => {
+            let workspace = require_workspace(&paths, &runtime.profile, workspace)?;
+            run_card(&client, &workspace, &agent_id)?;
+        }
+        AgentCommand::Send(args) => {
+            let workspace = require_workspace(&paths, &runtime.profile, args.workspace.clone())?;
+            run_send(&client, &workspace, args)?;
+        }
+        AgentCommand::Task { command } => {
+            let workspace =
+                require_workspace(&paths, &runtime.profile, task_workspace_flag(&command))?;
+            run_task(&client, &workspace, command)?;
+        }
     }
 
     Ok(())