Repository navigation
mcpx: adopt what is useful from codex-connectors-mcp #1984
Description
Relocated from dezren39/darwin-nix#346 on 2026-10-28 (placement audit): mcpx's single source lives in developing-today/code (pkgs/mcpx); re-home tracked at #309. Original body below; issue references inside it are historical (dezren39/nix numbering).
https://github.com/wileai/codex-connectors-mcp bridges the apps connected in Codex to MCP. Narrower than mcpx, but several of its choices are worth taking:
- mcpx: optional bearer token for /v1 and /mcp #1975 optional bearer token on /v1 and /mcp (was nix#338) — adopted, merged in mcpx: optional bearer-token auth for /v1 and /mcp (#1975) #2711 (
85960a65); child issue still open - mcpx: block further writes after a write whose outcome is unknown #1977 block writes after a write whose outcome is unknown (was nix#339) — adopted, merged in mcpx: guard unknown write outcomes (#1977) #4721 (
a41cba35); child issue still open - mcpx: a write policy -- ask, allow, deny -- with the exact call shown when asking #1979 write policy: ask / allow / deny, showing the exact call (was nix#340) — adopted, merged in mcpx: write policy gate (policy.write ask/allow/deny) — #1979 Phase 1 #2829 (
bd48deb1); child closed - mcpx: JSON Schema validation of tool inputs/outputs and elicitation schemas (7 conformance rows) #1951 JSON Schema validation of tool arguments, all three drafts, no coercion (was nix#254) — adopted, merged in mcpx: close #1951 remainder — stale dialect-gap rows + structuredContent warn-only policy doc #2190 (
178465d3); child closed - mcpx: a packaging test -- install the built artifact and check the MCP handshake #1981 a packaging test against the built artifact (was nix#341) — adopted, merged in test(mcpx): packaging test for the built artifact (#1981) #2994 (
0acf0caa); child closed - mcpx: front codex-connectors-mcp, for OAuth-authenticated apps without mcpx doing OAuth #1982 front codex-connectors-mcp itself, for OAuth-authenticated apps (was nix#342) — adopted, merged in feat(mcpx): add codex-connectors upstream — 37 connected Codex apps via the bridge (#1982) #2068 (
ebc66d78) fronting the bridge with 37 connected apps; child closed
Nothing in the list is rejected; the only rejected item is the comparison subject's model, below.
Not adopted: its one-upstream, one-account model and its stateful legacy sessions -- mcpx is a multi-upstream gateway serving every revision.
Comparison: it speaks MCP through the TypeScript SDK v1, so 2025-11-25 and earlier (initialize, Mcp-Session-Id), not 2026-07-28.
Acceptance criteria
- Every adoption checkbox resolves one way or the other: adopted items link their implementing PR/issue state, and rejected items say so in one line
- The tracker is closed when its last open child (fix(id): cut the interrupted sync session on committed state, not a record count (#188) #254 at the time of writing) resolves, or is explicitly handed to a named successor tracker — mcpx: JSON Schema validation of tool inputs/outputs and elicitation schemas (7 conformance rows) #1951 (was fix(id): cut the interrupted sync session on committed state, not a record count (#188) #254) is closed; last open children are mcpx: optional bearer token for /v1 and /mcp #1975 and mcpx: block further writes after a write whose outcome is unknown #1977, both implemented and merged but not yet closed
Activity
developing-today commentedon Oct 28, 2026
CODE-909
dezren39 commentedon Oct 28, 2026
History carried over from dezren39/darwin-nix#346 (verbatim, oldest first):
@dezren39 · 2026-10-27T11:10:39Z
Intake review — fleet monitor (2026-10-27)
Tracker status (checked this run): children #338, #339, #340, #341, #342 are all still OPEN, as is #254 (the JSON Schema validation item, which also gates #355). Nothing under this tracker has landed yet, so the tracker itself is fully live.
Note: the children are being enriched individually in the same monitor pass (#340, #341, #342 this run), each with its own acceptance criteria; this tracker stays the rollup.
Suggested approach: work the children in dependency order — #254 first (it gates #355 and underpins the argument-validation items), then #341's packaging test (it de-risks all the others' verification).
dezren39 commentedon Oct 28, 2026
🤖 Agent note · picked up
dezren39 commentedon Oct 28, 2026
🤖 Agent note · done
Triage verdict (task 2888): TRACKER CONFIRMED LIVE — all six children already relocated code-side; no duplicate, no new packet needed
This tracker arrived with nix-numbered children. Every one of them was relocated in the same 2026-10-28 placement audit and now lives here (checked fresh 2026-10-28T23:1xZ; the body checkboxes above now link the code numbers):
Dedupe / scope
pkgs/mcpxon main; duplicating that here would be re-work.Routing
No follow-up packet is enqueued for the tracker itself: a tracker is not implementable work, and each child already carries its own queued task in this queue (acceptance's "split with linked sub-issues" branch — the split exists and is now linked from the body). This issue stays open as the rollup and closes when its last open child resolves, per its own acceptance criteria.
18 remaining items
dezren39 commentedon Nov 3, 2026
agent-note:v1
state: triaged (coverage pass)
Coverage pass 2026-11-03: reviewed in the fleet-wide per-issue pass (drift / overlap / already-handled / stale / blockers / ordering). Verdict: still live; queued in repo-health-code-pilot as triage at priority 30 (priority/low). If this is already handled at current main, the triage closes it with evidence; if it needs splitting, the triage files the split.
Found via: coverage pass 2026-11-03
dezren39 commentedon Nov 3, 2026
🤖 Agent note · picked up
dezren39 commentedon Nov 4, 2026
🤖 Agent note · triage findings + closing (task 8571, lane code-h)
1c1aa4a79Verdict: tracker complete — closing. Fresh verification of every line: