feat: rehydrate MCP sessions from disk on codex-reply

[zsxkib]

motivation

i use codex as an MCP server inside claude code a lot. the two work really well as a team — claude handles the high-level orchestration and codex does the heavy lifting with shell access, file edits, etc.

the problem is when i've been working with just codex directly and it already has all the context for what i'm doing. i want to tell claude "go chat with codex 019c7616-b3e9-7782-9ea8-b53e6bb09329" and have it pick up right where i left off. but if the MCP server has restarted since that session (which happens all the time), codex-reply just returns "Session not found" — even though the full JSONL transcript is still sitting on disk at ~/.codex/sessions/.

this is frustrating because the data is RIGHT THERE. all the building blocks to fix it already exist in the codebase, they just aren't wired together in the error path.

what this does

adds a fallback in the codex-reply error path — if a thread isn't found in memory, it tries to load it from the persisted rollout file and resume the session with the full conversation history.

how it works

basically chains three functions that already existed but weren't connected:

• find_thread_path_by_id_str — locates the JSONL rollout file for the threadId
• read_session_meta_line — recovers the original cwd from session metadata
• resume_thread_from_rollout — replays the conversation history into a new thread

if no rollout file exists on disk, the original "Session not found" error is returned unchanged — so this is purely additive, no behavior change for the normal case.

the response already includes threadId in the output, so the client automatically picks up the new ID for subsequent calls.

files changed

file	change
codex-rs/core/src/thread_manager.rs	added pub fn auth_manager() getter (+4 lines)
codex-rs/mcp-server/src/message_processor.rs	added codex_home field, try_rehydrate_from_disk() method, modified error path (+75/-10 lines)

test plan

• cargo build -p codex-mcp-server — compiles
• cargo test -p codex-mcp-server — all 11 existing tests pass
• manual test: start session → restart MCP server → codex-reply with old threadId → session resumes from disk
• edge case: threadId with no JSONL file → still returns "Session not found"

[github-actions]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

Sakib Ahamed seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}

[etraut-openai]

We've updated our contribution guidelines to indicate that we're no longer accepting unsolicited code contributions. All code contributions are by invitation only. To read more about why we've taken this step, please refer to this announcement.