Skip to content

[Bug]: Codex MCP injection still describes launcher environment config #26120

Closed

Description

@coygeek

Summary

The Codex V2 provider-injection section in docs/orchestration-v2/orchestrator-mcp-server.md says Codex app-server receives the T3 MCP server through launcher -c overrides and T3_MCP_BEARER_TOKEN. Current adapter code instead supplies per-thread config with http_headers.Authorization in thread/start and thread/resume requests. The reference describes the wrong integration boundary for debugging provider setup.

Steps to reproduce

  1. In a checkout of each baseline under Version or commit, read the Codex V2 subsection under Provider Injection.
  2. In apps/server/src/orchestration-v2/Adapters/CodexAdapterV2.ts, inspect codexThreadRuntimeParams and its mcp_servers.t3-code config object.
  3. Follow that helper’s calls from ensureThread into thread/start and from resumeThread into thread/resume.
  4. Compare the request config with the document’s -c ...bearer_token_env_var example and statement about T3_MCP_BEARER_TOKEN.

These source/document comparisons were executed for all three baselines. No Codex process or provider session was launched.

Expected behavior

The existing repository documentation policy allows stale guidance to be rewritten or removed. Retained Codex integration guidance should identify the current per-thread request configuration, or point to the adapter instead of describing an obsolete launcher projection.

Actual behavior

For a defined MCP session, codexThreadRuntimeParams sets the T3 server URL, tool_timeout_sec to 3900, and http_headers.Authorization from mcpSession.authorizationHeader. ensureThread and resumeThread pass this config in their app-server requests. The documented launcher override and bearer environment-variable mechanism do not describe that active injection path.

This report establishes the documentation/source mismatch only. It does not claim authentication fails, that the environment variable can never exist through external configuration, or that the current header transport is defective.

Evidence

The reproduced failure is the documentation/source mismatch. The source branches explain the mismatch; runtime operations and project tests were not executed.

The orchestration-v2 document-set README describes target architecture. This report is limited to this page’s concrete present-tense description of implemented MCP behavior, and does not ask to implement the obsolete description. The repository documentation policy permits rewriting or removing stale guidance and linking to authoritative source.

Restoration check

Repeat the comparison against the adapter on the revision being reviewed. Passing documentation no longer identifies launcher -c overrides and the bearer environment variable as the adapter’s injection path; it accurately identifies or links to the per-thread start/resume configuration. No runtime transport or credential-handling change is requested.

Version or commit

Inspected repository source at current main 4914cdd98f3d15ab2ddef34f0af70d3d68a2d8e5, latest published nightly v0.1.4-nightly.20261027.3783 (dc84b0962efa4f4e05ef392c08986a9cd7c994d1), and latest stable v0.1.3 (d2c440312e5f39d59827ed466838a623516454d6). This is a platform-independent documentation/source comparison. No T3 server, provider session, live MCP mutation, or test suite was run.

Area

Docs

Impact

Cosmetic issue.

Workaround

Follow the adapter’s per-thread config helper and start/resume call sites, which were read directly in all inspected baselines.

Before submitting

  • I searched existing issues and did not find a duplicate.
  • I included enough detail to reproduce or investigate the problem.

References

Open report #15670 and open PR #23199 cover the tool count and recurring-task discoverability. They do not correct this contract. No comment is requested on either thread.

Triage assessment

  • Impact level: P3
  • Assessment status: supported
  • Impact basis: The verified defect is stale developer-reference wording about the integration boundary. No failed provider workflow or other consequential runtime effect was observed.
  • Workaround status: available
  • Workaround basis: Follow the adapter’s per-thread config helper and start/resume call sites, which were read directly in all inspected baselines.

Activity

  1. coygeek commented on Nov 9, 2026

    @coygeek
    Author

    Closing as resolved. Checked main at 6375070: the Codex V2 documentation now describes per-thread start/resume/fork configuration with an Authorization header, matching the adapter. The obsolete launcher -c and T3_MCP_BEARER_TOKEN guidance is gone, satisfying this issue’s documentation restoration check.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions