[codex] Use AgentAssertion downstream behind use_agent_identity by adrian-openai · Pull Request #17980 · openai/codex

adrian-openai · 2026-04-15T19:10:45Z

Summary

This is the AgentAssertion downstream slice for feature-gated agent identity support, replacing the oversized AgentAssertion slice from PR #17807.

It isolates task-scoped downstream AgentAssertion wiring on top of the merged PR3.1 work without re-carrying the earlier agent registration, task registration, or task-state history.

This PR includes the task-scoped bug-fix call sites from the review: generic file upload auth, MCP OpenAI file upload auth, and ARC monitor auth. Broader user/control-plane calls move to PR4.1 and PR4.2.

Stack

PR1: Add use_agent_identity feature flag #17385 - add features.use_agent_identity
PR2: Register agent identities behind use_agent_identity #17386 - register agent identities when enabled
PR3: Register agent tasks behind use_agent_identity #17387 - register agent tasks when enabled
PR3.1: Persist and prewarm agent tasks per thread #17978 - persist and prewarm registered tasks per thread
PR4: this PR - use task-scoped AgentAssertion downstream when enabled
PR4.1: [codex] Use background agent task auth for backend calls #18094 - introduce AuthManager-owned background/control-plane AgentAssertion auth
PR4.2: [codex] Use background task auth for additional backend calls #18260 - use background task auth for additional backend/control-plane calls

What Changed

add AgentAssertion envelope generation in codex-core
route downstream HTTP and websocket auth through AgentAssertion when an agent task is present
extend the model-provider auth provider so non-bearer authorization schemes can be passed through cleanly
make generic file uploads attach the full authorization header value
make MCP OpenAI file uploads use the cached thread agent task assertion when present
make ARC monitor calls use the cached thread agent task assertion when present

Why

The original PR had drifted ancestry and showed a much larger diff than the semantic change actually required. Restacking it onto PR3.1 keeps the reviewable surface down to the downstream assertion slice.

Validation

just fmt
cargo check -p codex-core -p codex-login -p codex-analytics -p codex-app-server -p codex-cloud-requirements -p codex-cloud-tasks -p codex-models-manager -p codex-chatgpt -p codex-model-provider -p codex-mcp -p codex-core-skills
cargo test -p codex-model-provider bearer_auth_provider
cargo test -p codex-core agent_assertion
cargo test -p codex-app-server remote_control
cargo test -p codex-cloud-requirements fetch_cloud_requirements
cargo test -p codex-models-manager manager::tests
cargo test -p codex-chatgpt
cargo test -p codex-cloud-tasks
cargo test -p codex-login agent_identity
just fix -p codex-core -p codex-login -p codex-analytics -p codex-app-server -p codex-cloud-requirements -p codex-cloud-tasks -p codex-models-manager -p codex-chatgpt -p codex-model-provider -p codex-mcp -p codex-core-skills
just fix -p codex-app-server
git diff --check

efrazer-oai

I had Codex go find all the places that are still hardcoded to use Bearer as discussed on Slack, here are the results:

As it stands, we have successfully migrated the shared Responses HTTP, Responses websocket, compact, and ChatGPT-backend realtime-call paths.

But there are still several other backend surfaces that build Authorization: Bearer ... directly or go through helpers that only understand bearer tokens. Those still need to migrate if we want this change to be exhaustive.

Shared backend client paths that still need the new auth shape:

codex-rs/backend-client/src/client.rs
- /wham/usage
- /wham/tasks/list
- /wham/tasks/{id}
- /wham/tasks/{id}/turns/{turn_id}/sibling_turns
- /wham/config/requirements
- POST /wham/tasks

Current callers of that shared client include:

codex-rs/cloud-requirements/src/lib.rs
codex-rs/app-server/src/codex_message_processor.rs
codex-rs/cloud-tasks-client/src/http.rs

Direct backend callers that still attach bearer auth themselves:

codex-rs/codex-api/src/files.rs
- /files
- /files/{id}/uploaded
codex-rs/core/src/mcp_openai_file.rs
codex-rs/models-manager/src/manager.rs
- /codex/models
codex-rs/codex-mcp/src/mcp/mod.rs
- .../wham/apps
codex-rs/chatgpt/src/chatgpt_client.rs
- shared GET helper used by:
  - codex-rs/chatgpt/src/get_task.rs
  - connector directory loading paths
codex-rs/core/src/connectors.rs
- connector directory requests
codex-rs/core/src/plugins/remote.rs
- /plugins/list
- /plugins/featured
- /plugins/{id}/enable
- /plugins/{id}/uninstall
codex-rs/core-skills/src/remote.rs
- /hazelnuts
- /hazelnuts/{skill_id}/export
codex-rs/analytics/src/client.rs
- /codex/analytics-events/events
codex-rs/core/src/arc_monitor.rs
- /codex/safety/arc
codex-rs/app-server/src/transport/remote_control/enroll.rs
- /wham/remote/control/server/enroll
codex-rs/app-server/src/transport/remote_control/websocket.rs
- websocket handshake for /wham/remote/control/server
codex-rs/cloud-tasks/src/util.rs
codex-rs/cloud-tasks/src/env_detect.rs
- /wham/environments
- /wham/environments/by-repo/...

One backend hit I found that does not look like part of this auth migration is:

codex-rs/core/src/plugins/startup_sync.rs
- backup archive fetch at /backend-api/plugins/export/curated

So the main point is: this PR fixes the shared model path, but it does not yet migrate all of the places where we talk to codex-backend.

nicksteele-oai

Looks clean to me

nicksteele-oai · 2026-04-16T16:02:14Z

+            .map_err(|error| format!("failed to read ChatGPT auth for file upload: {error}"))?;
+        CoreAuthProvider::from_bearer_token(Some(token_data.access_token), token_data.account_id)
    };
    let uploaded = upload_local_file(


Will file upload with ChatGPT auth be affected if the above code block fails?

## Summary Stack PR3 for feature-gated agent identity support. This PR adds per-thread agent task registration behind `features.use_agent_identity`. Tasks are minted on the first real user turn and cached in thread runtime state for later turns. ## Stack - PR1: #17385 - add `features.use_agent_identity` - PR2: #17386 - register agent identities when enabled - PR3: #17387 - this PR, original task registration slice - PR3.1: #17978 - persist and prewarm registered tasks per thread - PR4: #17980 - use `AgentAssertion` downstream when enabled ## Validation Covered as part of the local stack validation pass: - `just fmt` - `cargo test -p codex-core --lib agent_identity` - `cargo test -p codex-core --lib agent_assertion` - `cargo test -p codex-core --lib websocket_agent_task` - `cargo test -p codex-api api_bridge` - `cargo build -p codex-cli --bin codex` ## Notes The full local app-server E2E path is still being debugged after PR creation. The current branch stack is directionally ready for review while that follow-up continues.

shijie-oai · 2026-04-17T20:07:56Z

 use tokio::time::timeout;

-const DEFAULT_TIMEOUT: Duration = Duration::from_secs(10);
+const DEFAULT_TIMEOUT: Duration = Duration::from_secs(20);


hm maybe we should drop this? I assume it has to do with windows test flake?

Yes! The test is flaky locally as a result of the 10s.

shijie-oai · 2026-04-17T20:14:41Z

-    if let Some(token) = auth.bearer_token() {
-        request = request.bearer_auth(token);
+    if let Some(authorization) = auth.authorization_header_value() {
+        request = request.header(AUTHORIZATION, authorization);


This would enable us to use non bearer token (agent identity) for as the auth header vs bearer_auth is for bearer token only I assume.

shijie-oai · 2026-04-17T20:18:03Z

    let shell_command = vec![
-        "powershell".to_string(),
-        "-Command".to_string(),
-        "Start-Sleep -Seconds 1".to_string(),
+        "cmd.exe".to_string(),
+        "/d".to_string(),
+        "/c".to_string(),
+        "ping".to_string(),
+        "-n".to_string(),
+        "2".to_string(),
+        "127.0.0.1".to_string(),
    ];


No action: I think this is updated/different on latest main already.

shijie-oai · 2026-04-17T20:31:15Z

 pub struct ModelClientSession {
    client: ModelClient,
    websocket_session: WebsocketSession,
+    agent_task: Option<RegisteredAgentTask>,


Do we need this? agent_task is scoped to a session (thread) and here we are on the top level?

The field is in ModelClientSession,

The registered agent task is owned by the Codex thread/session state.

But each model turn needs to carry that task into all Responses requests made during that turn: HTTP stream, websocket prewarm, retries, continuation requests.

Putting it on ModelClient would be wrong because ModelClient is longer-lived and shared for the Codex session.

Passing it into every lower-level request method would be more error-prone.

Storing it on the per-turn ModelClientSession also lets the code avoid reusing a cached bearer-auth websocket when an AgentAssertion task is present.
Let me know if you disagree with that ^

shijie-oai · 2026-04-17T20:36:25Z

-    let token_data = auth
-        .get_token_data()
-        .map_err(|error| format!("failed to read ChatGPT auth for file upload: {error}"))?;
-    let upload_auth = CoreAuthProvider {
-        token: Some(token_data.access_token),
-        account_id: token_data.account_id,


Did not realize that we have so many locally built headers....

shijie-oai

Makes sense to me overall.

This was referenced Apr 15, 2026

Use AgentAssertion downstream behind use_agent_identity #17807

Closed

Register agent tasks behind use_agent_identity #17387

Merged

adrian-openai requested review from efrazer-oai, nicksteele-oai and shijie-oai April 15, 2026 21:04

adrian-openai marked this pull request as ready for review April 15, 2026 21:05

efrazer-oai reviewed Apr 15, 2026

View reviewed changes

adrian-openai force-pushed the dev/adrian/codex/agent-identity-use-assertion-split branch from ed316a9 to d1373d4 Compare April 15, 2026 23:53

adrian-openai changed the base branch from dev/adrian/codex/agent-identity-register-task to dev/adrian/codex/agent-task-state-prewarm April 15, 2026 23:53

adrian-openai force-pushed the dev/adrian/codex/agent-identity-use-assertion-split branch 2 times, most recently from dfd9aa0 to 688aad3 Compare April 16, 2026 01:11

github-actions Bot mentioned this pull request Apr 16, 2026

📊 AI CLI 工具社区动态日报 2026-04-16 gsscsd/big_model_radar#193

Open

adrian-openai mentioned this pull request Apr 16, 2026

[codex] Use background agent task auth for backend calls #18094

Merged

nicksteele-oai approved these changes Apr 16, 2026

View reviewed changes

nicksteele-oai reviewed Apr 16, 2026

View reviewed changes

adrian-openai mentioned this pull request Apr 16, 2026

Persist and prewarm agent tasks per thread #17978

Merged

adrian-openai force-pushed the dev/adrian/codex/agent-task-state-prewarm branch 2 times, most recently from ea5ec0a to 56f4b38 Compare April 17, 2026 03:35

adrian-openai mentioned this pull request Apr 17, 2026

[codex] Use background task auth for additional backend calls #18260

Merged

adrian-openai force-pushed the dev/adrian/codex/agent-task-state-prewarm branch 2 times, most recently from 3dcdb8d to 67cff31 Compare April 17, 2026 17:20

shijie-oai reviewed Apr 17, 2026

View reviewed changes

Comment thread codex-rs/core/src/agent_identity/assertion.rs Outdated

shijie-oai reviewed Apr 17, 2026

View reviewed changes

shijie-oai approved these changes Apr 17, 2026

View reviewed changes

adrian-openai force-pushed the dev/adrian/codex/agent-task-state-prewarm branch from e6f7ce3 to 0000662 Compare April 18, 2026 03:16

Base automatically changed from dev/adrian/codex/agent-task-state-prewarm to main April 19, 2026 22:45

adrian-openai force-pushed the dev/adrian/codex/agent-identity-use-assertion-split branch 2 times, most recently from 68a45b4 to c6d377f Compare April 20, 2026 01:10

adrian-openai added 4 commits April 19, 2026 21:49

Use AgentAssertion downstream behind use_agent_identity

7a230ec

Use agent assertions for task-scoped backend calls

1614e03

Preserve FedRAMP routing for assertion auth

a538ff9

Route task assertion auth through AuthManager

044222d

adrian-openai force-pushed the dev/adrian/codex/agent-identity-use-assertion-split branch from 2d2fd76 to 044222d Compare April 20, 2026 05:17

adrian-openai merged commit b44d285 into main Apr 20, 2026
34 of 36 checks passed

adrian-openai deleted the dev/adrian/codex/agent-identity-use-assertion-split branch April 20, 2026 06:16

github-actions Bot locked and limited conversation to collaborators Apr 20, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[codex] Use AgentAssertion downstream behind use_agent_identity#17980

[codex] Use AgentAssertion downstream behind use_agent_identity#17980
adrian-openai merged 4 commits intomainfrom
dev/adrian/codex/agent-identity-use-assertion-split

adrian-openai commented Apr 15, 2026 •

edited

Loading

Uh oh!

efrazer-oai left a comment

Uh oh!

nicksteele-oai left a comment

Uh oh!

nicksteele-oai Apr 16, 2026

Uh oh!

shijie-oai Apr 17, 2026

Uh oh!

adrian-openai Apr 17, 2026

Uh oh!

shijie-oai Apr 17, 2026

Uh oh!

shijie-oai Apr 17, 2026

Uh oh!

Uh oh!

shijie-oai Apr 17, 2026

Uh oh!

adrian-openai Apr 17, 2026

Uh oh!

shijie-oai Apr 17, 2026

Uh oh!

shijie-oai left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

adrian-openai commented Apr 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Stack

What Changed

Why

Validation

Uh oh!

efrazer-oai left a comment

Choose a reason for hiding this comment

Uh oh!

nicksteele-oai left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

shijie-oai left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

adrian-openai commented Apr 15, 2026 •

edited

Loading