fix(mcp): auto-fall-back to user-installed providers.json; backfill UNIFIED_PROVIDERS_CONFIG

[jobordu]

Summary

Closes the real root cause of the persistent unhealthy@1ms / error state in /nf:mcp-status that PRs #158 / #159 / #161 didn't fully address. Even on a fresh Claude Code session with all prior fixes pulled into main, every vanilla MCP slot still exits with Unknown PROVIDER_SLOT and never registers its tools.

Root cause

The mcpServers entries that /nf:mcp-setup historically created have only env: { PROVIDER_SLOT: <name> } — no UNIFIED_PROVIDERS_CONFIG. The MCP server defaults to __dirname/providers.json, which ships intentionally empty (populated at install time into ~/.claude/nf/bin/providers.json). So the spawned server sees zero providers, can't find its PROVIDER_SLOT, and exits before any tools register.

Fix

Two complementary changes:

1. bin/unified-mcp-server.mjs — when neither UNIFIED_PROVIDERS_CONFIG nor a populated repo-source providers.json is available, fall back to ~/.claude/nf/bin/providers.json. This self-heals existing user installs with zero install re-run required — the next session start picks up the right config.

Precedence: explicit env → populated repo source → user-installed copy.

2. bin/install.js — when creating mcpServers entries OR re-processing existing ones, set UNIFIED_PROVIDERS_CONFIG explicitly. Old entries get backfilled on next install so the env is self-documenting going forward.

Verified live

$ PROVIDER_SLOT=codex-1 node bin/unified-mcp-server.mjs
[mcp] Resolved codex: codex -> /Users/jonathanborduas/.nvm/.../bin/codex
[mcp] Resolved gemini: gemini -> /Users/jonathanborduas/.nvm/.../bin/gemini
[mcp] Resolved opencode: opencode -> /opt/homebrew/bin/opencode
[mcp] Resolved copilot: copilot -> /opt/homebrew/bin/copilot
[mcp] Resolved claude:  claude -> /Users/jonathanborduas/.local/bin/claude
[unified-mcp-server] started [slot: codex-1]
→ health_check returns: {"healthy":true,"latencyMs":38}

Previously: exited with Unknown PROVIDER_SLOT: codex-1 at session start.

Test plan

• npm run test:ci — 1550 pass, 0 fail
• npm run lint:isolation — clean
• npm run check:assets — clean
• Live: spawn server with only PROVIDER_SLOT env, no UNIFIED_PROVIDERS_CONFIG → starts cleanly, health_check returns healthy:true
• After merge + restart Claude Code: /nf:mcp-status shows all 7 slots available

🤖 Generated with Claude Code

[coderabbitai]

Warning

Rate limit exceeded

@jobordu has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 8 minutes and 49 seconds before requesting another review.

You’ve run out of usage credits. Purchase more in the billing tab.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 1dd26e81-c735-4291-8bef-20b17eeb6b35

📥 Commits

Reviewing files that changed from the base of the PR and between 600908a and e1e00fa.

📒 Files selected for processing (3)

• .secrets.baseline
• bin/install.js
• bin/unified-mcp-server.mjs

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/unified-mcp-fallback-config

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Copilot]

Pull request overview

This PR fixes MCP slot startup failures by ensuring unified-mcp-server can always find a populated providers.json, even for older MCP entries that only set PROVIDER_SLOT. It introduces a runtime fallback to the user-installed providers config and updates the installer to explicitly set/backfill UNIFIED_PROVIDERS_CONFIG in ~/.claude.json.

Changes:

• Add providers config resolution precedence in unified-mcp-server.mjs (env override → repo source if populated → user-installed copy).
• Backfill UNIFIED_PROVIDERS_CONFIG into existing ~/.claude.json mcpServers entries during install, and set it for newly-created entries.
• Refresh .secrets.baseline line references due to code movement.

Reviewed changes

Copilot reviewed 1 out of 3 changed files in this pull request and generated no comments.

File	Description
bin/unified-mcp-server.mjs	Adds config loading logic to fall back to the user-installed providers.json when the repo copy is empty/unset.
bin/install.js	Ensures MCP server entries always include UNIFIED_PROVIDERS_CONFIG, including backfilling older entries.
.secrets.baseline	Updates stored line numbers / generated timestamp after code shifts.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.