Merge Unified's GitHub integration Merge Agent Handler's GitHub MCP server

Connect your AI Agents to GitHub in minutes

Merge lets you securely connect your agents to GitHub and thousands of tools instantly

Start building for free

Read docs

GitHub

Learn more about Merge Agent Handler

Available tools

list_workflows

List all workflows in a repository's .github/workflows directory. Returns workflow IDs needed for other Actions tools

get_workflow

Get a specific workflow by ID or filename (e.g. 'ci.yml'). Use list_workflows to discover workflow IDs and filenames

list_workflow_runs

List workflow runs for a repository or specific workflow. Filter by branch, status, actor, event, head_sha, or created date range. Returns run history with status and conclusion

get_workflow_run

Get details for a specific workflow run including status, conclusion, and timing. Use listworkflowruns to find run IDs

trigger_workflow

Trigger a workflow dispatch event to manually run a workflow (it must have a workflowdispatch trigger). Returns the new run's ID/URL when the API provides them; if absent, find the run via listworkflowruns with workflowid, branch=ref, and event='workflow_dispatch'

list_workflow_jobs

List jobs for a workflow run. Each job represents a step in the workflow. Use getjoblogs to retrieve logs for a specific job

get_job_logs

Get log content for a workflow job. Returns the last 500 lines by default; set taillines=0 for the full log, or maxbytes to cap the size. Use listworkflowjobs to find job IDs

cancel_workflow_run

Cancel a workflow run that is queued or in progress. Use listworkflowruns to find run IDs

rerun_workflow

Re-run all jobs in a workflow run, optionally with debug logging enabled. Use listworkflowruns to find run IDs; to retry only failures use rerunfailedjobs

rerun_failed_jobs

Re-run only the failed jobs of a workflow run (and their dependents), optionally with debug logging enabled. Use listworkflowruns to find run IDs

list_workflow_run_artifacts

List artifacts produced by a workflow run, optionally filtered by exact artifact name. Use listworkflowruns to find run IDs; pass artifact IDs to getartifactdownload_url

get_artifact_download_url

Get a short-lived download URL for a workflow run artifact's zip archive (the URL expires after about 1 minute). Use listworkflowrun_artifacts to find artifact IDs

list_deployments

List a repository's deployments, optionally filtered by commit SHA, ref or environment. Returns deployment IDs for listdeploymentstatuses, which is where the outcome of each deployment lives. A deployment record on its own does not say whether it succeeded

list_deployment_statuses

List the statuses recorded against one deployment, newest first. The first entry is the deployment's current state (success, failure, inprogress and so on). Use listdeployments to find deployment IDs

list_environment_secrets

List the names of secrets configured on a repository environment. Names and timestamps only: GitHub never returns secret values through the API at any scope, so this cannot read a secret. Use it to check which secrets an environment expects

list_environments

List the deployment environments configured on a repository. Use this to find valid environment names before calling listenvironmentsecrets or filtering list_deployments by environment; guessing a name returns the same 404 as a missing permission, which is hard to tell apart. Empty is normal for a repository with no environments

get_repository_archive_url

Get a short-lived download URL for a repository archive (zip or tar) at a branch, tag or commit. Returns the URL rather than the bytes, because a repository archive is far too large to pass through an agent's context. The URL is signed and expires within minutes, so use it promptly

list_discussion_categories

List discussion categories in a repository. Call this first to get category node IDs and slugs for filtering list_discussions. Categories are also required when creating discussions

list_discussions

List discussions in a repository with server-side filters: category (node ID or slug, use listdiscussioncategories first), answered, state (OPEN/CLOSED), and sort by created/updated time. Cursor pagination via limit/cursor

get_discussion

Get a discussion by number including body, category, author, comment count, and answer info. Returns nodeid, which creatediscussioncomment uses internally. Use listdiscussions to find discussion numbers

get_discussion_comments

Get a discussion's comments and their threaded replies, flattened with each reply after its parent (replytonodeid links them). Each comment includes its nodeid, needed by updatediscussioncomment and as replytocommentid. limit/cursor paginate top-level comments only. Use listdiscussions to find discussion numbers

create_discussion

Create a discussion in a repository. Call listdiscussioncategories first to get a valid category slug or node ID. Discussions must be enabled on the repo (set hasdiscussions via updaterepository). Returns the new discussion including its number for creatediscussioncomment

create_discussion_comment

Add a comment to a discussion by number. Supports GitHub Markdown. Set replytocommentid (a comment nodeid from getdiscussioncomments) to reply in a thread. Returns the new comment including its node_id for later updates

update_discussion_comment

Update the body of an existing discussion comment. commentid is the comment node ID (e.g. 'DC...') returned by getdiscussioncomments or creatediscussioncomment

get_file_contents

Get file or directory contents from a repository. Files return base64 content and metadata; files over 1MB are re-fetched as raw text (encoding 'raw'). If size exceeds maxsizebytes (default 1MB), metadata is returned with content omitted and a message. Directories return a list of entries. Use ref for branch/tag/SHA

create_or_update_file

Create or update a single file in a repository. Provide sha of existing file to update it (get sha from getfilecontents). Content is plain UTF-8 text by default; set encoding='base64' if you are passing already-encoded content. Omitting branch commits to the repository's DEFAULT branch

delete_file

Delete a file from a repository. Requires the file's current blob SHA (get sha from getfilecontents). Cannot delete directories. Omitting branch commits the deletion to the repository's DEFAULT branch

push_files

Push multiple files in a single commit using the Git Data API (atomic). Per-file content is plain UTF-8 text by default; set encoding='base64' on a file if you are passing already-encoded content. Omitting branch commits to the repository's DEFAULT branch. More efficient than createorupdate_file for multiple files

get_repository_tree

Get the file tree for a repository. Set recursive=true to include all files in all subdirectories in one call; on very large repositories the response may be cut off — truncated=true means partial results, so fetch subtrees individually by SHA instead. tree_sha can be a branch name, tag, or commit SHA

get_blob

Read a file's raw content by its blob SHA. Use this when getrepositorytree returns truncated=true and you must fetch subtrees or files individually, or when you already hold a blob SHA. To read by path instead, use getfilecontents. Returns decoded text; binary blobs set is_binary=true and leave content null

list_gists

List gists for the authenticated user or a specific user. Gists are public or secret code snippets stored on GitHub

get_gist

Get a specific gist by ID including all file contents

create_gist

Create a new gist with one or more files. Set public=true for a public gist visible to everyone, or public=false for a secret gist accessible only via its URL

update_gist

Update a gist's description or file contents. To delete a file, set its value to null in the files map. Only the gist owner can update it

delete_gist

Delete a gist permanently. Only the gist owner can delete it

get_issue

Get a single issue by number from a specific repository. Returns labels, assignees, milestone, and a comment count (use getissuecomments for comment bodies), plus the issue 'id' needed by sub-issue tools

get_issues

Get a list of issues from a repository with filtering by state, labels, assignee, creator, and more. Supports pagination and sorting. Note: GitHub interleaves pull requests in this list; items with a 'pull_request' field are PRs — skip them when only issues are wanted

search_issues

Search for issues across all accessible repositories using GitHub's search syntax. Supports advanced filtering and sorting options

create_issue

Create a new issue with title, body, labels, assignees, milestone, and type. Milestone numbers come from listmilestones, assignable usernames from listassignees, issue type names from listissuetypes, label names from list_labels. Returns the created issue including its number and id

update_issue

Update an issue's title, body, state, labels, assignees, milestone, or type. labels and assignees REPLACE the full existing set. Set removemilestone/removeassignees true to clear them. Milestone numbers: listmilestones; usernames: listassignees; type names: listissuetypes. To mark a duplicate, set statereason 'duplicate' and duplicateofissueid to the canonical issue's numeric id from get_issue (an id, not an issue number)

close_issue

Close an issue by setting its state to closed. Optionally specify a statereason (completed, notplanned, duplicate); when statereason is 'duplicate', pass duplicateofissueid with the canonical issue's numeric id from get_issue (an id, not an issue number)

reopen_issue

Reopen a closed issue by setting its state to open

get_issue_comments

Get comments for a specific issue. Supports filtering by date and pagination for issues with many comments

create_issue_comment

Create a new comment on an issue. Comments support GitHub markdown formatting

update_issue_comment

Update an existing issue comment's content. Only the comment author or repository owners can update comments

delete_issue_comment

Delete an issue comment. Only the comment author or repository owners can delete comments

add_labels_to_issue

Add labels to an issue. Labels must already exist in the repository — use listlabels to browse them or createlabel to add new ones

remove_label_from_issue

Remove a specific label from an issue

set_issue_labels

Set labels on an issue, replacing all existing labels with the new set

remove_all_labels_from_issue

Remove all labels from an issue

lock_issue

Lock an issue to prevent new comments. Only collaborators will be able to add new comments

unlock_issue

Unlock an issue to allow new comments from all users

list_labels

List labels for a repository with pagination

get_label

Get a single repository label by name. Use list_labels to browse all labels

create_label

Create a new label in a repository. Color must be a 6-character hex code without the # (e.g. 'f29513')

update_label

Update an existing repository label's name, color, or description. Use list_labels to find label names

delete_label

Delete a label from a repository. This removes the label from all issues and pull requests that have it

list_milestones

List a repository's milestones with state, sort, and pagination options. Use the returned milestone 'number' as the milestone value in createissue/updateissue

list_assignees

List users who can be assigned to issues in a repository. Use the returned 'login' values as assignee/assignees in createissue/updateissue

list_issue_types

List an organization's issue types (e.g. Bug, Task). Use the returned 'name' values as the type in createissue/updateissue. Only available for organization-owned repositories

list_sub_issues

List an issue's sub-issues with pagination. Returns full issue objects including the 'id' needed by addsubissue/removesubissue

add_sub_issue

Add an existing issue as a sub-issue of a parent issue. subissueid is the child issue's ID, NOT its number — getissue returns it as 'id'. Set replaceparent true to re-parent a sub-issue that already has a parent. Returns the parent issue

remove_sub_issue

Remove a sub-issue from its parent issue. subissueid is the child issue's ID, NOT its number — getissue or listsub_issues return it as 'id'. The child issue itself is not deleted. Returns the parent issue

list_user_issues

List issues across ALL repositories accessible to the authenticated user (assigned/created/mentioned/subscribed), unlike getissues which is scoped to one repository. Each item's 'repository' field identifies its repo; items with a 'pullrequest' field are PRs

list_notifications

List notifications for the authenticated user. Set all=true to include read notifications. Filter to a specific repo by providing owner and repo. Paginated: per_page max 100 (the API defaults to 50 when pagination is omitted)

get_notification

Get details for a specific notification thread by ID. Use list_notifications to find thread IDs

mark_notification_read

Mark a notification thread as read. The thread stays in the inbox, marked read. Use marknotificationdone to remove it from the inbox, or dismiss_notification to stop receiving future notifications for it

mark_all_notifications_read

Mark all notifications as read. Scope to one repository by providing BOTH owner and repo; providing only one is rejected. With neither, the ENTIRE account's notifications are marked read

mark_notification_done

Mark a notification thread as done, removing it from the inbox. Unlike marknotificationread (thread stays in the inbox as read) and dismissnotification (unsubscribes from future notifications). Use listnotifications to find thread IDs

dismiss_notification

Unsubscribe from a notification thread so you no longer receive future notifications for it. Does not mark existing notifications read or done. Use marknotificationread or marknotificationdone for those

manage_notification_subscription

Mute or unmute a notification thread. Set ignored=true to mute the thread (suppresses all future notifications), ignored=false to unmute. The thread endpoint only supports muting; to fully unsubscribe from a thread use dismiss_notification

manage_repo_notification_subscription

Manage repository-level notification subscription. Set subscribed=true to watch all notifications, ignored=true to ignore all notifications for this repository

get_pull_request

Get a single pull request by number from a specific repository. Returns detailed PR information including merge status, reviews, and file changes

get_pull_requests

Get a list of pull requests from a repository with filtering options by state, head/base branches, and sorting. Supports pagination

create_pull_request

Create a pull request from head into base. Provide title, or issue (an issue number) to convert an existing issue into a PR. Use headrepo for cross-repository PRs in the same network. Supports draft PRs; to assign reviewers, call requestreviewers after creation

update_pull_request

Update an existing pull request's title, body, state, base branch, and other properties

merge_pull_request

Merge a pull request using the specified merge method (merge, squash, or rebase). Fails if the PR is not mergeable; check mergeable via getpullrequest first if unsure

close_pull_request

Close a pull request without merging it. The PR can be reopened later if needed

reopen_pull_request

Reopen a closed pull request. The PR must not be merged to be reopened

convert_to_draft

Convert a pull request to draft status, preventing it from being merged until marked ready for review (markreadyfor_review)

mark_ready_for_review

Mark a draft pull request as ready for review, allowing it to be merged once approved. Use converttodraft to move it back to draft

get_pull_request_reviews

Get reviews for a pull request including approval status, comments, and reviewer information

create_pull_request_review

Create a pull request review. Set event to APPROVE, REQUESTCHANGES, or COMMENT to submit immediately; omit event to create a PENDING review, then publish it with submitpullrequestreview or discard it with deletependingreview

submit_pull_request_review

Submit a PENDING pull request review with APPROVE, REQUESTCHANGES, or COMMENT. Use getpullrequestreviews (state PENDING) or the ID returned by createpullrequest_review to find review IDs

dismiss_pull_request_review

Dismiss a submitted pull request review with a required message explaining why. Use getpullrequestreviews to find review IDs. Only submitted reviews can be dismissed; delete PENDING reviews with deletepending_review

delete_pending_review

Delete a PENDING (unsubmitted) pull request review and its draft comments. Use getpullrequestreviews (state PENDING) to find review IDs. Submitted reviews cannot be deleted; use dismisspullrequestreview instead

create_pull_request_comment

Create a review comment on a pull request for code review. Comments specific lines of code in pull request diffs

request_reviewers

Request specific users or teams to review a pull request. Reviewers will be notified of the request

remove_review_request

Remove review requests from specific users or teams for a pull request

get_pull_request_files

Get the files changed in a pull request with additions, deletions, and change counts. Set includepatch=true for per-file diffs, or use getpullrequestdiff for the whole-PR diff

check_if_merged

Check if a pull request has been merged and get merge details if applicable

get_pull_request_commits

Get commits in a pull request showing the individual changes that make up the PR. Supports pagination (up to 250 commits are listable); use has_more to fetch further pages

update_pull_request_branch

Update the head branch of a pull request to include the latest changes from the base branch

get_commit_status

Get the combined status for a specific commit or ref (legacy statuses API). Returns a concise summary: state, per-context statuses, and counts

get_check_runs

Get check runs for a specific commit or ref (Checks API). Filter by checkname or status, and choose latest or all attempts via filter. Paginated (perpage max 100); has_more signals further pages

get_check_suites

List the check suites for a commit or ref. A suite is one CI provider's whole run; getcheckruns returns the individual checks across all suites, and getchecksuiteruns narrows to one suite. Filter by appid to see a single provider. Paginated (perpage max 100); hasmore signals further pages

get_check_suite_runs

List the check runs inside one check suite. Use getchecksuites to find the suite ID. Prefer getcheckruns when you want every check on a commit regardless of provider; use this to inspect a single provider's suite. Paginated (perpage max 100); hasmore signals further pages

get_check_run_annotations

Get a check run's annotations: the file, line range and message for each problem the check reported. This is how a failing check says WHICH line broke, which getcheckruns does not include. Use getcheckruns or getchecksuiteruns to find the checkrun_id. An empty list means the check published no annotations, not an error

get_pull_request_checks

Get an overall CI status summary for a PR's head commit, including checks and legacy status

add_reply_to_pull_request_comment

Reply to an existing review comment on a pull request. Use listpullrequestreviewthreads (database_id on each comment) to find review comment IDs

get_pull_request_diff

Get the unified diff for a pull request as plain text. Large diffs are truncated at maxbytes (default 200000 bytes); truncated=true signals cut-off content. Use getpullrequestfiles for per-file stats

list_pull_request_review_threads

List review threads on a pull request with resolution status, file path, and comments. Returns thread node IDs for resolvereviewthread/unresolvereviewthread and comment databaseids for addreplytopullrequestcomment. Paginate by passing end_cursor as cursor

resolve_review_thread

Resolve a pull request review thread, marking the conversation as addressed. Requires the thread node ID returned by listpullrequestreviewthreads

unresolve_review_thread

Reopen (unresolve) a resolved pull request review thread. Requires the thread node ID returned by listpullrequestreviewthreads

get_repository

Get a single repository by owner and name. Returns detailed repository information including stats, settings, and metadata

get_repositories

List ONLY the authenticated user's repositories (owned, collaborator, or org member). Filter by type/visibility/affiliation, sort, and paginate. To find repositories owned by other users or orgs, use search_repositories with org: or user: qualifiers

create_repository

Create a new repository. Can create private or public repositories with various settings like issues, wiki, etc

create_organization_repository

Create a new repository in an organization. Can create private or public repositories with various settings like issues, wiki, etc

update_repository

Update an existing repository's settings, description, visibility, and other properties

delete_repository

Delete a repository permanently. This action cannot be undone. Use with extreme caution

get_branches

Get repository branches with optional filtering by protection status. Includes commit information for each branch

create_branch

Create a new branch from an existing commit. Requires the SHA of the commit to branch from

get_commits

Get repository commits with filtering options by author, date range, file path, and branch/SHA

get_commit

Get a single commit by SHA with detailed information including file changes, statistics, and parent commits. For large commits, the files list is paginated: up to 300 files per page (use page/per_page), 3,000 files max in total

get_releases

Get repository releases including tags, assets, and release notes. Returns both published and draft releases

get_latest_release

Get the latest release for a repository. Returns the most recent non-prerelease, non-draft release. Raises a not-found error if the repository has no releases; use get_releases to list drafts and prereleases

check_starred

Check if the authenticated user has starred a repository

star_repository

Star a repository for the authenticated user

unstar_repository

Unstar a repository for the authenticated user

fork_repository

Fork a repository into the authenticated user's account or a specified organization. Forking is asynchronous: the fork's metadata is returned immediately, but its git data can take several minutes to provision, so reads/pushes against the fork may fail or look empty at first. Retry after a short wait

list_repository_collaborators

List collaborators for a repository. Filter by affiliation (outside, direct, all) and by minimum permission (pull, triage, push, maintain, admin)

get_release_by_tag

Get a release by its tag name (e.g. 'v1.0.0'). Use get_releases to list all releases

list_tags

List tags for a repository with pagination

search_orgs

Search for organizations on GitHub. Use qualifiers like location:, repos:>N, followers:>N. Sort by followers, repositories, or joined; supports pagination. 'type:org' is added automatically unless the query already has a type: qualifier

create_release

Create a release. tagname is required: use an existing tag from listtags, or a new tag name (created from targetcommitish, default branch if omitted). Set generatereleasenotes to auto-fill notes. Returns the release id, htmlurl, and upload_url. Binary asset upload is not supported

update_release

Update a release by releaseid (from getreleases, getreleasebytag, or createrelease). Only supplied fields change: tagname, name, body, draft, prerelease, makelatest, target_commitish

delete_release

Delete a release by releaseid (from getreleases or getreleaseby_tag). Deletes only the release; the underlying git tag remains

compare_commits

Compare two commits, branches, or tags. basehead uses 'BASE...HEAD' syntax, e.g. 'main...my-feature' or 'v1.0.0...v2.0.0' (cross-fork: 'owner:branch'). Returns aheadby/behindby/status, the commits in range, and changed files. Paginate with pagination.page/per_page; files are capped at 300 per page

get_branch_protection

Get effective branch protection: classic branch protection, falling back to repository and organization rulesets. protectionsource says which applies ('classic', 'ruleset', 'none'); protected=false only when neither does. Use before mergepullrequest to explain why a merge is blocked. Use getbranches to list branch names

search_repositories

Search GitHub repositories using search syntax. Supports qualifiers like language:python, stars:>100, topic:machine-learning, user:octocat, org:github

search_code

Search code across GitHub repositories. Must include a qualifier: repo:, org:, user:, language:, path:, or extension:. Example: 'addClass in:file language:js repo:jquery/jquery'. Results include text_matches snippets showing the matching fragments, so you rarely need to fetch each file

search_commits

Search commits across GitHub repositories. Use qualifiers like author:, committer:, author-date:, repo:, org:. Example: 'fix bug author:octocat'

search_users

Search GitHub users and organizations. Use qualifiers like type:user, type:org, location:, language:, followers:>N, repos:>N. Example: 'tom repos:>42 followers:>1000'

search_pull_requests

Search pull requests across GitHub repositories. Supports filtering by state, author, reviewer, label, base branch. Example: 'is:open review-requested:octocat base:main'

get_authenticated_user

Get the authenticated user's profile information including public and private details

get_user

Get detailed information about any GitHub user by their username

get_organization

Get detailed information about a GitHub organization by its name

get_followers

Get a list of users who follow a specific user. Supports pagination for users with many followers

get_following

Get a list of users that a specific user follows. Supports pagination for users who follow many people

check_following

Check whether one user follows another user. Set targetuser to the user who may be followed. Omit username to check the authenticated (connected) user, or set it to check whether some other user follows targetuser

follow_user

Follow a user as the authenticated user. The user will be notified of the follow. Requires the user:follow scope

unfollow_user

Unfollow a user as the authenticated user. Requires the user:follow scope

get_organization_members

Get members of an organization with filtering options by role and 2FA status

get_user_organizations

Get organizations that a user is publicly a member of

get_organization_teams

Get teams in an organization. Requires organization membership or public visibility

get_team_members

Get members of a specific team with filtering by role (member or maintainer)

check_team_membership

Check a user's membership in a team, including its state (active vs pending invitation) and role

add_team_membership

Add a user to a team with specified role (member or maintainer). Requires admin permissions

remove_team_membership

Remove a user from a team. Requires admin permissions

create_team

Create a team in an organization. Requires org admin permissions. permission accepts only pull or push; grant finer roles (triage, maintain, admin) per-repository afterwards

get_user_memberships

Get the authenticated user's organization memberships including private memberships. Filter by state (active or pending) and paginate for users in many organizations

get_user_events

Get public events for a user showing their recent GitHub activity. Supports pagination

get_user_starred_repos

Get repositories starred by a user, sorted by star time (created) or last push (updated), with pagination

validate_credential

Validate GitHub credentials. Verifies credentials during setup.

View all tools by creating a free account See more tools

How to set up Merge Agent Handler

In an mcp.json file, add the configuration below, and restart Cursor.

Learn more in the official documentation ↗

1{
2  "mcpServers": {
3    "agent-handler": {
4      "url": "https://ah-api-develop.merge.dev/api/v1/tool-packs/{TOOL_PACK_ID}/registered-users/{REGISTERED_USER_ID}/mcp",
5      "headers": {
6        "Authorization": "Bearer yMt*****"
7      }
8    }
9  }
10}
11

Copy Code

Open your Claude Desktop configuration file and add the server configuration below. You'll also need to restart the application for the changes to take effect.

Make sure Claude is using the Node v20+.

Learn more in the official documentation ↗

1{
2  "mcpServers": {
3    "agent-handler": {
4      "command": "npx",
5      "args": [
6        "-y",
7        "mcp-remote@latest",
8        "https://ah-api-develop.merge.dev/api/v1/tool-packs/{TOOL_PACK_ID}/registered-users/{REGISTERED_USER_ID}/mcp",
9        "--header",
10        "Authorization: Bearer ${AUTH_TOKEN}"
11      ],
12      "env": {
13        "AUTH_TOKEN": "yMt*****"
14      }
15    }
16  }
17}

Copy Code

Open your Windsurf MCP configuration file and add the server configuration below.
Click on the refresh button in the top right of the Manage MCP server page or in the top right of the chat box in the box icon.

Learn more in the official documentation ↗

1{
2    "mcpServers": {
3      "agent-handler": {
4        "command": "npx",
5        "args": [
6          "-y",
7          "mcp-remote@latest",
8          "https://ah-api.merge.dev/api/v1/tool-packs/<tool-pack-id>/registered-users/<registered-user-id>/mcp",
9          "--header",
10          "Authorization: Bearer ${AUTH_TOKEN}"
11        ],
12        "env": {
13          "AUTH_TOKEN": "<ah-production-access-key>"
14        }
15      }
16    }
17  }

Copy Code

In Command Palette (Cmd+Shift+P on macOS, Ctrl+Shift+P on Windows), run "MCP: Open User Configuration".

You can then add the configuration below and press "start" right under servers. Enter the auth token when prompted.

Learn more in the official documentation ↗

1{
2  "inputs": [
3    {
4      "type": "promptString",
5      "id": "agent-handler-auth",
6      "description": "Agent Handler AUTH_TOKEN", // "yMt*****" when prompt
7      "password": true
8    }
9  ],
10  "servers": {
11    "agent-handler": {
12      "type": "stdio",
13      "command": "npx",
14      "args": [
15        "-y",
16        "mcp-remote@latest",
17        "https://ah-api-develop.merge.dev/api/v1/tool-packs/{TOOL_PACK_ID}/registered-users/{REGISTERED_USER_ID}/mcp",
18        "--header",
19        "Authorization: Bearer ${input:agent-handler-auth}"
20      ]
21    }
22  }
23}

Copy Code

FAQs on using Merge's GitHub MCP server

How can I use the GitHub MCP server?

Here are just a few use cases worth implementing:

Monitor pull request activity across repositories. Your agent can track all open pull requests across multiple repositories and send daily summaries to your team's Slack channel, highlighting PRs that need review, are blocked, or have been open longer than your team's SLA

Automate code review coordination. Your agent can analyze PR content, automatically request reviews from the appropriate team members based on file paths and code ownership, and send reminders to reviewers when PRs are waiting for feedback

Link development work to business objectives. Your agent can combine GitHub commit and PR data with your project management tool to show leadership how engineering work maps to product roadmap items, helping them understand resource allocation and velocity toward strategic goals

Streamline issue management from communication tools. Your agent can help users create, update, and track GitHub issues directly from Slack or Microsoft Teams, allowing developers to triage bugs, assign tasks, and update statuses without context-switching between tools

What are popular tools for GitHub’s MCP server?

Here are some popular tools across data types:

Issues

<code class="blog_inline-code">get_issues</code>
<code class="blog_inline-code">search_issues</code>
<code class="blog_inline-code">close_issue</code>
<code class="blog_inline-code">add_labels_to_issue</code>

Pull requests (PRs)

<code class="blog_inline-code">get_pull_requests</code>
<code class="blog_inline-code">create_pull_requests</code>
<code class="blog_inline-code">update_pull_requests</code>
<code class="blog_inline-code">reopen_pull_request</code>

Repositories

<code class="blog_inline-code">get_repository</code>
<code class="blog_inline-code">create_repository</code>
<code class="blog_inline-code">update_repository</code>
<code class="blog_inline-code">delete_repository</code>

What makes Merge Agent Handler’s GitHub MCP server better than alternative GitHub MCP servers?

Merge Agent Handler provides several platform-level advantages over standalone MCP servers:

Enterprise-grade security and DLP: All GitHub tool inputs and outputs are scanned by Merge Agent Handler’s Security Gateway, which can block, redact, or mask sensitive data based on configurable rules. This allows you to enforce data governance policies consistently across all agent interactions

Managed authentication and credential handling: Merge Agent Handler supports guided authentication flows, credential storage, and both individual and shared authentication models, removing the need to implement OAuth or token handling yourself

Real-time observability and audit trails: Every GitHub tool call is logged with a complete, fully-searchable audit trail, enabling debugging, compliance reviews, and performance optimization from a single dashboard

Tool Pack management and customization: GitHub tools can be bundled with other connectors into Tool Packs, allowing you to control exactly which tools an agent can access per use case, without changing agent logic

Can I set custom security rules for GitHub tool calls in Merge Agent Handler?

Yes, you can set custom security rules for any GitHub tool calls to block or redact sensitive data. You can also log tool calls, allowing the agent to make them while providing your team with real-time visibility whenever this happens.

Here are some specific examples:

Block your agents from creating or updating pull requests, issues, or commits with API keys, access tokens, or credentials in titles, descriptions, comments, or code content

Automatically redact sensitive data patterns like internal IP addresses, database connection strings, or proprietary algorithms when your agents retrieve or display code, PR descriptions, or issue details, so they only see masked versions

Monitor and log whenever your agents attempt to push code or create PRs containing sensitive terms like "password", "secret", or "private key" in file names or content, giving your security team visibility without blocking the action

Block your agents from creating GitHub issues or PR descriptions with confidential project codenames, customer names under NDA, or internal-only labels that shouldn't be visible in your repositories

How can I start using Merge Agent Handler’s GitHub MCP server?

Here are all the steps you’ll need to take to integrate your agent with the MCP server:

1. Create a free Agent Handler account. Sign up here to access the Agent Handler dashboard.

2. Create a Tool Pack. Tool Packs define which connectors and tools your agent can access. This involves navigating to Tool Packs in the dashboard, clicking Create Tool Pack, adding the GitHub connector, and choosing your authentication mode (Individual or Shared).

3. Set up a Registered User. Go to Registered Users in the dashboard; each Registered User represents a person or system identity that your agent runs as. When a user needs to authenticate with GitHub, Merge Agent Handler will automatically display a guided authorization prompt—no custom OAuth code required.

4. Test in the Playground. Before integrating into your app, validate everything works. This means opening the Sandbox for your Tool Pack and testing GitHub tool calls.

5. Integrate with your application. Connect your agent to Agent Handler to access the tools. You can do this in two ways:

Via MCP config: Copy the MCP server configuration from your Tool Pack page and add it to your agent's MCP client settings. This gives your agent access to all configured tools through the Model Context Protocol
Via API: Use Agent Handler's REST API to list available tools and invoke them programmatically. You'll need your Agent Handler API key, Tool Pack ID, and Registered User ID to authenticate requests