Files
coder/agent/proto/agent.proto
T
Kyle Carberry cd3692c0c2 feat: add agent-side workspace context sources and Agent API v2.10 PushContextState (#25983)
Adds the agent half of the workspace context sources RFC. The agent now
resolves instruction files, skills, and MCP configs into a typed
`Snapshot`, watches the relevant paths recursively, exposes the source
list over a workspace-agent HTTP API, and pushes each `Snapshot` to
coderd over a new `PushContextState` RPC on Agent API v2.10.

The coderd-side handler is a stub returning `Unimplemented` for now.
Real persistence to `workspace_agent_context`, chatd hydration on dirty
events, and the `KindMCPServer` MCP provider are tracked by
[CODAGT-569](https://linear.app/codercom/issue/CODAGT-569/enable-agent-api-v210-pushcontextstate-bump-currentminor-wire-coderd).
This matches the pattern used for v2.7 `ReportBoundaryLogs` in
[#21293](https://github.com/coder/coder/pull/21293), which bumped the
version and shipped a stub server so the wire and client could iterate
before the persistence layer landed.

## What ships

### agent/agentcontext (new package)

- `Source`, `Resource` (kinds `instruction_file`, `skill`, `mcp_config`,
`mcp_server` plus reserved `plugin`/`hook`/`subagent`/`command`),
`ResourceStatus`, `Snapshot`, `ComputeAggregateHash`.
- `Manager` owns the in-memory source list, performs the initial resolve
synchronously in `NewManager`, runs a re-resolve/watcher loop in `Run`,
exposes
`AddSource`/`RemoveSource`/`Sources`/`HasSource`/`Snapshot`/`SubscribeChanges`/`Resync`/`SeedSources`/`Close`.
- `Resolver` walks scan roots, classifies recognized files, enforces 64
KiB per-resource, 2 MiB aggregate, and 500-resource caps with
`StatusOversize`/`StatusExcluded`/`StatusUnreadable`/`StatusInvalid`
outcomes, skips `node_modules`/`vendor`/etc., validates symlink targets
stay inside the scan root, stamps `SourcePath` on user-derived
resources, and optionally pulls MCP server tool lists via an
`MCPProvider` interface. MCP config resources ship metadata only (size,
hash) so secrets in env blocks never leave the agent.
- `Watcher` is a recursive `fsnotify` wrapper with a 250 ms debounce,
dynamic arming of newly created directories, and an ENOSPC-tolerant
degraded mode that no-ops further syncs until the manager resyncs
explicitly.
- HTTP API for `GET/POST /sources`, `GET/DELETE /sources/{path}`, `POST
/resync` mounted at `/api/v0/context`.
- `Pusher` interface plus `RunPush` goroutine with exponential backoff
capped at 30 s. `DRPCPusher` adapts the generated `DRPCAgentClient210`
to `Pusher` and translates `drpcerr.Unimplemented` to
`ErrPushUnimplemented` so the push loop exits cleanly when talking to
coderd deployments that have not enabled the real handler.

### agent/proto (v2.10)

- New messages `ContextResource`, `PushContextStateRequest`,
`PushContextStateResponse` and the `PushContextState` RPC on `service
Agent`.
- Generated `DRPCAgentClient210` interface and
`codersdk/agentsdk.Client.ConnectRPC210` / `ConnectRPC210WithRole`.
- `tailnet/proto.CurrentMinor` bumped from `9` to `10`.

### Agent wiring

- `agent.Options.Client` declares both v2.9 and v2.10 connectors;
`run()` dials with `ConnectRPC210WithRole`.
- `apiConnRoutineManager` holds a `DRPCAgentClient210`. Existing v2.8
routines keep their narrower `DRPCAgentClient28` signature thanks to
interface embedding.
- `startAgentAPI210` is the v2.10 counterpart to `startAgentAPI` for
routines that need the new client. The push context state routine uses
it.
- A `contextManager` is constructed in `agent.init()`, seeded from the
existing `CODER_AGENT_EXP_*_DIRS` env vars, started in its own goroutine
under `gracefulCtx`, and closed in `agent.Close`.
- `handleManifest` calls `Manager.SeedSources` for sources rooted at the
manifest directory, then `Resync` after `manifest.Swap`, so the snapshot
reflects the workspace working directory immediately instead of waiting
for the next filesystem event.
- HTTP routes mounted at `/api/v0/context` when the manager is up.

### Coderd stub

`coderd/agentapi/context.go` returns `drpcerr.Unimplemented` for
`PushContextState`. The real handler that persists
`workspace_agent_context` rows, hydrates chats, and emits dirty events
lives in CODAGT-569.

## Tests

24 tests across `agent/agentcontext` cover types, paths, resolver
behavior with file caps, skill containers, MCP secret omission, symlink
target validation, the recursive watcher firing on real fsnotify events,
manager source CRUD / `Resync` / `SeedSources` / `Run` lifetime, the
HTTP API, the DRPC adapter, and the push retry / initial-flag /
unimplemented paths. Passes `go test -race -count=2`.

`TestAgent_ContextStatePushed` boots a full agent against
`agenttest.FakeAgentAPI` (which now records `PushContextState` traffic)
and asserts the seeded `AGENTS.md` appears in a snapshot push with
`schema_version = 1`.

<details>
<summary>Notes for reviewers</summary>

- Source CRUD is workspace-agent-token only; coderd is not in the path
for source mutation.
- Per-resource cap 64 KiB, aggregate 2 MiB, count cap 500; resources
past the cap ship with `StatusExcluded` and an empty payload so the
aggregate hash still detects content edits. MCP-emitted resources
enforce both a per-provider count cap and the aggregate byte cap.
- Symlinks inside the scan root are followed; symlinks pointing outside
(or broken) are rejected with `StatusExcluded` so credentials reachable
via a stray symlink stay off the wire.
- The initial push gates `lifecycle = ready` in the eventual full
design. For this PR the `SeedSources` plus `handleManifest`-driven
`Resync` keeps the snapshot fresh; the live push loop ships now and
DRPCPusher translates the coderd `Unimplemented` stub into a clean exit.
- The `PLUGIN`/`HOOK`/`SUBAGENT`/`COMMAND` kinds are reserved in proto
and Go enums but unused; the Claude Code plugin resolver ships in a
follow-up that does not need a schema migration.
- Two follow-ups remain, both tracked by
[CODAGT-569](https://linear.app/codercom/issue/CODAGT-569/enable-agent-api-v210-pushcontextstate-bump-currentminor-wire-coderd):
(1) the chatd-side handler that persists snapshots and dirties chats;
(2) the `coder exp chat context` CLI command set for
`list`/`show`/`add`/`remove`/`refresh`.

</details>

_This PR was authored by Coder Agents on Kyle Carberry's behalf._
2026-06-08 12:08:40 -07:00

675 lines
16 KiB
Protocol Buffer

syntax = "proto3";
option go_package = "github.com/coder/coder/v2/agent/proto";
package coder.agent.v2;
import "tailnet/proto/tailnet.proto";
import "google/protobuf/timestamp.proto";
import "google/protobuf/duration.proto";
import "google/protobuf/empty.proto";
import "google/protobuf/struct.proto";
message WorkspaceApp {
bytes id = 1;
string url = 2;
bool external = 3;
string slug = 4;
string display_name = 5;
string command = 6;
string icon = 7;
bool subdomain = 8;
string subdomain_name = 9;
enum SharingLevel {
SHARING_LEVEL_UNSPECIFIED = 0;
OWNER = 1;
AUTHENTICATED = 2;
PUBLIC = 3;
ORGANIZATION = 4;
}
SharingLevel sharing_level = 10;
message Healthcheck {
string url = 1;
google.protobuf.Duration interval = 2;
int32 threshold = 3;
}
Healthcheck healthcheck = 11;
enum Health {
HEALTH_UNSPECIFIED = 0;
DISABLED = 1;
INITIALIZING = 2;
HEALTHY = 3;
UNHEALTHY = 4;
}
Health health = 12;
bool hidden = 13;
}
message WorkspaceAgentScript {
bytes log_source_id = 1;
string log_path = 2;
string script = 3;
string cron = 4;
bool run_on_start = 5;
bool run_on_stop = 6;
bool start_blocks_login = 7;
google.protobuf.Duration timeout = 8;
string display_name = 9;
bytes id = 10;
}
message WorkspaceAgentMetadata {
message Result {
google.protobuf.Timestamp collected_at = 1;
int64 age = 2;
string value = 3;
string error = 4;
}
Result result = 1;
message Description {
string display_name = 1;
string key = 2;
string script = 3;
google.protobuf.Duration interval = 4;
google.protobuf.Duration timeout = 5;
}
Description description = 2;
}
message Manifest {
bytes agent_id = 1;
string agent_name = 15;
string owner_username = 13;
bytes workspace_id = 14;
string workspace_name = 16;
uint32 git_auth_configs = 2;
map<string, string> environment_variables = 3;
string directory = 4;
string vs_code_port_proxy_uri = 5;
string motd_path = 6;
bool disable_direct_connections = 7;
bool derp_force_websockets = 8;
optional bytes parent_id = 18;
coder.tailnet.v2.DERPMap derp_map = 9;
repeated WorkspaceAgentScript scripts = 10;
repeated WorkspaceApp apps = 11;
repeated WorkspaceAgentMetadata.Description metadata = 12;
repeated WorkspaceAgentDevcontainer devcontainers = 17;
repeated WorkspaceSecret secrets = 19;
}
// WorkspaceSecret is a secret included in the agent manifest
// for injection into a workspace.
message WorkspaceSecret {
// Environment variable name to inject (e.g. "GITHUB_TOKEN").
// Empty string means this secret is not injected as an env var.
string env_name = 1;
// File path to write the secret value to (e.g.
// "~/.aws/credentials"). Empty string means this secret is not
// written to a file.
string file_path = 2;
// The decrypted secret value.
bytes value = 3;
}
message WorkspaceAgentDevcontainer {
bytes id = 1;
string workspace_folder = 2;
string config_path = 3;
string name = 4;
optional bytes subagent_id = 5;
}
message GetManifestRequest {}
message ServiceBanner {
bool enabled = 1;
string message = 2;
string background_color = 3;
}
message GetServiceBannerRequest {}
message Stats {
// ConnectionsByProto is a count of connections by protocol.
map<string, int64> connections_by_proto = 1;
// ConnectionCount is the number of connections received by an agent.
int64 connection_count = 2;
// ConnectionMedianLatencyMS is the median latency of all connections in milliseconds.
double connection_median_latency_ms = 3;
// RxPackets is the number of received packets.
int64 rx_packets = 4;
// RxBytes is the number of received bytes.
int64 rx_bytes = 5;
// TxPackets is the number of transmitted bytes.
int64 tx_packets = 6;
// TxBytes is the number of transmitted bytes.
int64 tx_bytes = 7;
// SessionCountVSCode is the number of connections received by an agent
// that are from our VS Code extension.
int64 session_count_vscode = 8;
// SessionCountJetBrains is the number of connections received by an agent
// that are from our JetBrains extension.
int64 session_count_jetbrains = 9;
// SessionCountReconnectingPTY is the number of connections received by an agent
// that are from the reconnecting web terminal.
int64 session_count_reconnecting_pty = 10;
// SessionCountSSH is the number of connections received by an agent
// that are normal, non-tagged SSH sessions.
int64 session_count_ssh = 11;
message Metric {
string name = 1;
enum Type {
TYPE_UNSPECIFIED = 0;
COUNTER = 1;
GAUGE = 2;
}
Type type = 2;
double value = 3;
message Label {
string name = 1;
string value = 2;
}
repeated Label labels = 4;
}
repeated Metric metrics = 12;
}
message UpdateStatsRequest{
Stats stats = 1;
}
message UpdateStatsResponse {
google.protobuf.Duration report_interval = 1;
}
message Lifecycle {
enum State {
STATE_UNSPECIFIED = 0;
CREATED = 1;
STARTING = 2;
START_TIMEOUT = 3;
START_ERROR = 4;
READY = 5;
SHUTTING_DOWN = 6;
SHUTDOWN_TIMEOUT = 7;
SHUTDOWN_ERROR = 8;
OFF = 9;
}
State state = 1;
google.protobuf.Timestamp changed_at = 2;
}
message UpdateLifecycleRequest {
Lifecycle lifecycle = 1;
}
enum AppHealth {
APP_HEALTH_UNSPECIFIED = 0;
DISABLED = 1;
INITIALIZING = 2;
HEALTHY = 3;
UNHEALTHY = 4;
}
message BatchUpdateAppHealthRequest {
message HealthUpdate {
bytes id = 1;
AppHealth health = 2;
}
repeated HealthUpdate updates = 1;
}
message BatchUpdateAppHealthResponse {}
message Startup {
string version = 1;
string expanded_directory = 2;
enum Subsystem {
SUBSYSTEM_UNSPECIFIED = 0;
ENVBOX = 1;
ENVBUILDER = 2;
EXECTRACE = 3;
}
repeated Subsystem subsystems = 3;
}
message UpdateStartupRequest{
Startup startup = 1;
}
message Metadata {
string key = 1;
WorkspaceAgentMetadata.Result result = 2;
}
message BatchUpdateMetadataRequest {
repeated Metadata metadata = 2;
}
message BatchUpdateMetadataResponse {}
message Log {
google.protobuf.Timestamp created_at = 1;
string output = 2;
enum Level {
LEVEL_UNSPECIFIED = 0;
TRACE = 1;
DEBUG = 2;
INFO = 3;
WARN = 4;
ERROR = 5;
}
Level level = 3;
}
message BatchCreateLogsRequest {
bytes log_source_id = 1;
repeated Log logs = 2;
}
message BatchCreateLogsResponse {
bool log_limit_exceeded = 1;
}
message GetAnnouncementBannersRequest {}
message GetAnnouncementBannersResponse {
repeated BannerConfig announcement_banners = 1;
}
message BannerConfig {
bool enabled = 1;
string message = 2;
string background_color = 3;
}
message WorkspaceAgentScriptCompletedRequest {
Timing timing = 1;
}
message WorkspaceAgentScriptCompletedResponse {
}
message Timing {
bytes script_id = 1;
google.protobuf.Timestamp start = 2;
google.protobuf.Timestamp end = 3;
int32 exit_code = 4;
enum Stage {
START = 0;
STOP = 1;
CRON = 2;
}
Stage stage = 5;
enum Status {
OK = 0;
EXIT_FAILURE = 1;
TIMED_OUT = 2;
PIPES_LEFT_OPEN = 3;
}
Status status = 6;
}
message GetResourcesMonitoringConfigurationRequest {
}
message GetResourcesMonitoringConfigurationResponse {
message Config {
int32 num_datapoints = 1;
int32 collection_interval_seconds = 2;
}
Config config = 1;
message Memory {
bool enabled = 1;
}
optional Memory memory = 2;
message Volume {
bool enabled = 1;
string path = 2;
}
repeated Volume volumes = 3;
}
message PushResourcesMonitoringUsageRequest {
message Datapoint {
message MemoryUsage {
int64 used = 1;
int64 total = 2;
}
message VolumeUsage {
string volume = 1;
int64 used = 2;
int64 total = 3;
}
google.protobuf.Timestamp collected_at = 1;
optional MemoryUsage memory = 2;
repeated VolumeUsage volumes = 3;
}
repeated Datapoint datapoints = 1;
}
message PushResourcesMonitoringUsageResponse {
}
message Connection {
enum Action {
ACTION_UNSPECIFIED = 0;
CONNECT = 1;
DISCONNECT = 2;
}
enum Type {
TYPE_UNSPECIFIED = 0;
SSH = 1;
VSCODE = 2;
JETBRAINS = 3;
RECONNECTING_PTY = 4;
}
bytes id = 1;
Action action = 2;
Type type = 3;
google.protobuf.Timestamp timestamp = 4;
string ip = 5;
int32 status_code = 6;
optional string reason = 7;
}
message ReportConnectionRequest {
Connection connection = 1;
}
message SubAgent {
string name = 1;
bytes id = 2;
bytes auth_token = 3;
}
message CreateSubAgentRequest {
string name = 1;
string directory = 2;
string architecture = 3;
string operating_system = 4;
message App {
message Healthcheck {
int32 interval = 1;
int32 threshold = 2;
string url = 3;
}
enum OpenIn {
SLIM_WINDOW = 0;
TAB = 1;
}
enum SharingLevel {
OWNER = 0;
AUTHENTICATED = 1;
PUBLIC = 2;
ORGANIZATION = 3;
}
string slug = 1;
optional string command = 2;
optional string display_name = 3;
optional bool external = 4;
optional string group = 5;
optional Healthcheck healthcheck = 6;
optional bool hidden = 7;
optional string icon = 8;
optional OpenIn open_in = 9;
optional int32 order = 10;
optional SharingLevel share = 11;
optional bool subdomain = 12;
optional string url = 13;
}
repeated App apps = 5;
enum DisplayApp {
VSCODE = 0;
VSCODE_INSIDERS = 1;
WEB_TERMINAL = 2;
SSH_HELPER = 3;
PORT_FORWARDING_HELPER = 4;
}
repeated DisplayApp display_apps = 6;
optional bytes id = 7;
}
message CreateSubAgentResponse {
message AppCreationError {
int32 index = 1;
optional string field = 2;
string error = 3;
}
SubAgent agent = 1;
repeated AppCreationError app_creation_errors = 2;
}
message DeleteSubAgentRequest {
bytes id = 1;
}
message DeleteSubAgentResponse {}
message ListSubAgentsRequest {}
message ListSubAgentsResponse {
repeated SubAgent agents = 1;
}
// BoundaryLog represents a log for a single resource access processed
// by boundary.
message BoundaryLog {
message HttpRequest {
string method = 1;
string url = 2;
// The rule that resulted in this HTTP request being allowed. Only populated
// when allowed = true because boundary denies requests by default and
// requires rule(s) that allow requests.
string matched_rule = 3;
}
// Whether boundary allowed this resource access.
bool allowed = 1;
// The timestamp when boundary processed this resource access.
google.protobuf.Timestamp time = 2;
// The resource being accessed by boundary.
oneof resource {
HttpRequest http_request = 3;
}
// Monotonically increasing integer assigned by boundary, starting at 0
// per session. Primary ordering key when boundary is in use.
int32 sequence_number = 4;
}
// ReportBoundaryLogsRequest is a request to re-emit the given BoundaryLogs.
message ReportBoundaryLogsRequest {
repeated BoundaryLog logs = 1;
// session_id identifies the boundary invocation that produced these
// logs. It is a UUID generated by boundary at startup and is the same
// for all batches produced by a single boundary run.
string session_id = 2;
// confined_process is the name of the process that boundary is
// confining (e.g. "claude-code", "codex", "copilot").
string confined_process_name = 3;
}
message ReportBoundaryLogsResponse {}
// UpdateAppStatusRequest updates the given Workspace App's status. c.f. agentsdk.PatchAppStatus
message UpdateAppStatusRequest {
string slug = 1;
enum AppStatusState {
WORKING = 0;
IDLE = 1;
COMPLETE = 2;
FAILURE = 3;
}
AppStatusState state = 2;
string message = 3;
string uri = 4;
}
message UpdateAppStatusResponse {}
// ContextResource is a single resolved workspace context
// resource (instruction file, skill meta, MCP config, or live
// MCP server tool list) pushed from the agent to coderd as part
// of a PushContextStateRequest snapshot.
//
// The resource kind is conveyed by which variant of the body
// oneof is set. Reserved variants for the Claude Code plugin
// RFC (plugin/hook/subagent/command bodies) are not emitted by
// v2.10 agents but will be added without renumbering.
message ContextResource {
// source is the resource's own locator: a canonical file path
// for file-backed kinds, or the MCP server name for
// mcp_server resources.
string source = 1;
// source_path is the user-declared scan root that produced
// this resource (empty for built-in roots, set to the owning
// .mcp.json for mcp_server entries declared in a user config).
optional string source_path = 2;
// content_hash is sha256 over the original on-disk bytes (or
// over the agent's canonical encoding for non-file kinds).
bytes content_hash = 3;
// size_bytes is the resource's original size in bytes.
uint64 size_bytes = 4;
Status status = 5;
// error carries the per-resource failure string when status
// is not OK; may also carry a non-fatal warning when status
// is OK.
string error = 6;
enum Status {
STATUS_UNSPECIFIED = 0;
OK = 1;
OVERSIZE = 2;
UNREADABLE = 3;
INVALID = 4;
EXCLUDED = 5;
}
// body conveys both the resource kind (via which variant is
// set) and the kind-specific payload. The variant is set even
// when status is not OK so coderd can still attribute the
// failure to a known kind.
oneof body {
InstructionFileBody instruction_file = 10;
SkillMetaBody skill = 11;
MCPConfigBody mcp_config = 12;
MCPServerBody mcp_server = 13;
}
// Reserved tags from the legacy v2.10 schema that carried
// id (1->renamed), kind enum, payload, description, and the
// removed plugin/hook/subagent/command flat fields. Keep them
// reserved so a future renumber cannot reintroduce them.
reserved 7, 8, 9, 14, 15, 16;
}
// InstructionFileBody carries a plain-text instruction file
// such as AGENTS.md, CLAUDE.md, or .cursorrules. The content is
// the verbatim file bytes (capped at the resolver's per-resource
// limit).
message InstructionFileBody {
bytes content = 1;
}
// SkillMetaBody carries the SKILL.md meta file content plus the
// fields parsed from its YAML front-matter. Supporting files in
// the skill directory are NOT included; clients fetch them on
// demand via the agent's local HTTP API.
message SkillMetaBody {
bytes meta = 1;
string name = 2;
string description = 3;
}
// MCPConfigBody is intentionally empty: the .mcp.json content
// can contain secrets in env blocks and must not leave the
// agent. content_hash and size_bytes on ContextResource still
// let coderd detect changes for cache invalidation.
message MCPConfigBody {
}
// MCPServerBody carries a live MCP server's resolved tool list,
// emitted by the agent's MCPProvider after the server has been
// connected.
message MCPServerBody {
string server_name = 1;
string description = 2;
repeated MCPTool tools = 3;
}
// MCPTool mirrors the MCP server-reported tool surface. The
// input schema is JSON Schema; we ship it as a google.protobuf
// Struct so coderd can introspect it without re-parsing JSON.
message MCPTool {
string name = 1;
string description = 2;
google.protobuf.Struct input_schema = 3;
}
message PushContextStateRequest {
uint64 version = 1;
bytes aggregate_hash = 2;
repeated ContextResource resources = 3;
bool initial = 4;
uint64 schema_version = 5;
string snapshot_error = 6;
}
message PushContextStateResponse {
bool accepted = 1;
}
service Agent {
rpc GetManifest(GetManifestRequest) returns (Manifest);
rpc GetServiceBanner(GetServiceBannerRequest) returns (ServiceBanner);
rpc UpdateStats(UpdateStatsRequest) returns (UpdateStatsResponse);
rpc UpdateLifecycle(UpdateLifecycleRequest) returns (Lifecycle);
rpc BatchUpdateAppHealths(BatchUpdateAppHealthRequest) returns (BatchUpdateAppHealthResponse);
rpc UpdateStartup(UpdateStartupRequest) returns (Startup);
rpc BatchUpdateMetadata(BatchUpdateMetadataRequest) returns (BatchUpdateMetadataResponse);
rpc BatchCreateLogs(BatchCreateLogsRequest) returns (BatchCreateLogsResponse);
rpc GetAnnouncementBanners(GetAnnouncementBannersRequest) returns (GetAnnouncementBannersResponse);
rpc ScriptCompleted(WorkspaceAgentScriptCompletedRequest) returns (WorkspaceAgentScriptCompletedResponse);
rpc GetResourcesMonitoringConfiguration(GetResourcesMonitoringConfigurationRequest) returns (GetResourcesMonitoringConfigurationResponse);
rpc PushResourcesMonitoringUsage(PushResourcesMonitoringUsageRequest) returns (PushResourcesMonitoringUsageResponse);
rpc ReportConnection(ReportConnectionRequest) returns (google.protobuf.Empty);
rpc CreateSubAgent(CreateSubAgentRequest) returns (CreateSubAgentResponse);
rpc DeleteSubAgent(DeleteSubAgentRequest) returns (DeleteSubAgentResponse);
rpc ListSubAgents(ListSubAgentsRequest) returns (ListSubAgentsResponse);
rpc ReportBoundaryLogs(ReportBoundaryLogsRequest) returns (ReportBoundaryLogsResponse);
rpc UpdateAppStatus(UpdateAppStatusRequest) returns (UpdateAppStatusResponse);
rpc PushContextState(PushContextStateRequest) returns (PushContextStateResponse);
}