Repository navigation
Conversation
Agent users need a single setup command that connects hooks and the running MCP companion. The existing hook installer selects every supported agent, and quickstart can report success while the companion is inactive. Use Kit's setup report so each integration can explain its own recovery. Keep Kit's released configuration behavior until its bearer preservation change ships.
Agent setup needs to find the same daemon and install hooks for the configuration the user chose. Recovery commands must retain that choice, including paths with spaces. When the daemon is stopped, a manual setup command cannot be generated yet. Report the reason and the next step so print mode does not silently produce no output. Config errors can have several causes, so reserve the data directory instruction for a relative path.
Keep coverage at the Forge integration boundary so dependency behavior does not obscure regressions in discovery, guidance, and hook setup. Exercise malformed configuration through the actual loader and check the order of daemon setup steps at their source.
Agent setup must use the selected configuration for discovery and hooks, and guide users to the right repair when it cannot load that configuration.
Print mode needs to explain how to connect when an agent cannot use the required token reference. An empty result leaves the user without a command or a next step. The setup check already supplies guidance for failed and unknown states, so its empty-step fallback is unnecessary.
Agent setup must retain user settings while updating Forge's token reference. Removing the old reference when authentication is disabled also lets clients connect without a stale credential.
Hooks need to reach the daemon selected during setup even when an agent starts with a different environment. Manual MCP setup also needs the token location so users can authenticate outside Forge-launched terminals. Keep recovery instructions tied to the selected configuration and point users at the underlying setup error.
Agent setup tests should catch mistakes in Forge's guidance and token selection without duplicating checks owned by the setup library. Keeping distinct daemon and authentication cases makes those regressions visible while reducing redundant fixtures.
The hook command already owns its flags, and its install options have only one caller. Keeping those definitions together makes the selected daemon settings easier to follow without changing how hooks find it.
Contributor
Author
|
This draft waits for a Kit release containing kenn-io/kit#368, which supplies the MCP settings preservation used here. |
roborev: Combined Review (
|
Agent setup prints a token read command for users who authenticate outside a Forge terminal. PowerShell interprets apostrophes and brackets in paths as syntax or wildcard patterns, so the command must preserve the path as a literal.
A first-time user can run setup before the daemon creates its configuration. Report how to start the daemon in that case, while keeping check mode free of writes. Keep hook commands consistent with existing installations and preserve user credentials when authentication changes. Setup uses the default configuration selected by the environment.
Keep coverage for daemon discovery and token selection while removing cases that repeat the same Forge behavior or exercise the setup library's own contracts.
A mistyped config path left installed hooks unable to reach Forge and made MCP discovery report a stopped daemon. Reject that path before installing hooks or reporting listener status. A fresh installation can still use the default path before the daemon creates its config.
Agent hooks need to reach the same daemon when the agent starts with a different environment. Setup also needs to honor a custom config and report MCP status when hook configuration fails. A missing default config is normal on a fresh install, while a missing custom config signals a selection error.
An upgrade should keep trusted hook commands when the configuration already selects the daemon's data directory. Configurations that depend on the home directory still need that selection carried into the hook's environment. Setup checks need to identify the token required for the next attempt, and quickstart guidance must retain a custom configuration path.
Repeated setup cases obscured which failures each test could catch. Keep the distinct config and guidance paths covered while making a missing default config check exercise hook discovery.
Agent setup needs to retain selected agents and the existing trusted hook command while giving users the daemon's next step. Keep the config file check for hook homes because runtime sources report the winning environment override and treat an empty file value as explicit.
The lint check rejects the existing condition. Preserve the exception for a missing default config while keeping the check clear to the linter.
roborev: Combined Review (
|
cpcloud
pushed a commit
to kenn-io/kit
that referenced
this pull request
Nov 14, 2026
…368) Users who tune an app's MCP server in Claude Code or Codex keep those settings when the app reinstalls it with env values, headers, or a bearer reference. Before, an app that supplied `BearerTokenEnvVar` as `APP_TOKEN` replaced the whole entry, so a user's `User='keep'` header, `tool_timeout_sec=20`, and Codex `enabled=false` disappeared. Now they stay. Before, rerunning setup with only the URL after turning off authentication left `bearer_token_env_var='APP_TOKEN'` in place, and Codex refused to connect while `APP_TOKEN` was unset. An app now names that variable in the new `Server.PreviousBearerTokenEnvVar` field, and setup removes it. When the app omits authentication, setup keeps any bearer setting the app doesn't name. App values win for the env and header keys the app supplies, and keys it stops supplying stay until the user removes them. Args follow the app, so a hand-added flag such as `--write` makes Check report the server as outdated. Check uses the same merge as Install. Needed by kenn-io/docbank#1491 Needed by kenn-io/forge#1997 Co-authored-by: Rod Boev <rodboev@users.noreply.github.com>
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Claude Code and Codex users can connect to Forge's hooks and MCP with one command. Today they run
agent-hook install, which without--agentwrites hooks for all 11 agents, and then copy the MCP URL frommcp quickstartinto each client by hand.Setup installs hooks for the installed Claude Code and Codex clients and registers the running daemon's MCP URL with both. When the API requires auth, the entry references
KENN_FORGE_API_TOKENinstead of the token, and outside Forge-launched terminals, setup tells you to export it from the reportedtoken_path. When setup can't finish, you see:failed: Forge daemon isn't runningand nameskenn-forge daemon start.--checkexits non-zero.data_dir, the hooks line fails and asks for an absolute one.--printgives the manual MCP commands, including the token export when it's missing.--config.mcp quickstartdrops its leadingOKwhen MCP is inactive and points toagent setup.When the config doesn't set its own
data_dir, hooks now pass--home, so an agent started later with a differentKENN_FORGE_HOMEstill reports to the daemon chosen at setup. This needs kenn-io/kit#368 released.go.modpinsgo.kenn.io/kitto a pseudo-version of that branch, so bump it to the release once it ships.