DevBridge MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CODEX_HOME | No | Optional Codex history location. Auto-detected by default. | |
| CONTEXT7_API_KEY | No | Context7 credential. Default is empty. | |
| BRIDGE_WRITE_ROOTS | Yes | Allowed project directories that the bridge may access, such as D:\Projects. The supplied .env.example uses an example placeholder; edit it to match your machine. Do not set broad drive roots unless you understand the risks. | Example placeholder; edit |
| BRIDGE_BROWSER_ENABLED | No | Enable browser testing. Set to 1 to enable. Default is off. | |
| BRIDGE_SEMGREP_ENABLED | No | Enable Semgrep scan. Set to 1 to enable. Default is off. Semgrep must be separately installed. | |
| BRIDGE_ACTIVITY_ENABLED | No | Enable local activity tracking. Set to 1 to enable. Default is off. | |
| BRIDGE_CONTEXT7_ENABLED | No | Enable Context7 documentation lookup. Set to 1 to enable. Default is off. | |
| BRIDGE_TERMINAL_ENABLED | No | Enable controlled terminal operations. Set to 1 to enable. Default is off. | |
| MACBRIDGE_MCP_TOKEN_FILE | No | Path to the MCP token file. Used for troubleshooting 503 MCP authentication unavailable; the bridge account must be able to read it. | |
| BRIDGE_BROWSER_ALLOWED_ORIGINS | No | Approved browser origins. Default is empty. | |
| BRIDGE_PROJECT_COMMANDS_ENABLED | No | Enable project build/test tasks. Set to 1 to enable. Default is off. | |
| BRIDGE_TERMINAL_SCRIPTS_ENABLED | No | Enable controlled PowerShell scripts. Set to 1 to enable. Default is off. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| scan_codex_historyA | Scan local Codex session files into the local search index. |
| search_codex_historyC | Search indexed Codex conversations. |
| latest_codex_sessionsC | Return recent Codex sessions. |
| get_codex_sessionB | Return messages from one Codex session. |
| codex_storage_infoC | Show detected Codex storage. |
| list_drivesB | List readable local Windows drives. |
| list_directoryB | List entries in an absolute local directory. |
| read_fileC | Read a local text file by absolute path. |
| file_infoC | Read metadata for an absolute local path. |
| search_filesA | Search text recursively from an absolute local directory. |
| write_fileB | Create or replace a UTF-8 text file inside an allowlisted development folder. |
| patch_fileB | Replace exact text inside a UTF-8 file in an allowlisted development folder. |
| browser_statusB | Check the opt-in Playwright browser configuration. |
| browser_openA | Open an approved HTTP(S) site in the dedicated Chrome profile; otherwise return local PowerShell approval instructions. Never self-approves websites. |
| browser_snapshotA | Read the current browser page title, URL, and visible text. |
| browser_screenshotB | Capture a viewport screenshot (may contain private ERPNext data). |
| browser_clickA | Click an exact locator. Can change ERP data; requires explicit authorization and confirm=true. |
| browser_fillC | Fill one exact locator. May trigger autosave; requires explicit authorization and confirm=true. |
| browser_source_probeA | Reload the authenticated development page and inspect only invalid resource references in the original HTML; excludes boot data. |
| browser_diagnosticsB | Inspect recent browser console errors, uncaught exceptions, HTTP errors and failed requests. Optional clear resets diagnostic history. |
| browser_closeB | Close the dedicated Playwright browser session. |
| terminal_planB | Plan a supported development command without executing it. |
| terminal_runC | Execute a controlled project command. PowerShell scripts need separate local opt-in and explicit confirmation. |
| project_task_planB | Inspect a defined npm build/test/lint task without executing it. |
| project_task_runA | Run a predefined npm build/test/lint task in an approved local project and return the exit code and logs. Requires BRIDGE_PROJECT_COMMANDS_ENABLED=1. |
| integration_catalogA | List available and pending development-tool integrations. |
| git_inspectB | Read local repository status and diff statistics (never stages or commits). |
| semgrep_scanB | Opt-in Semgrep static security scan; requires local Semgrep CLI and BRIDGE_SEMGREP_ENABLED=1. |
| browser_accessibility_basicA | Read-only heuristic accessibility checks in the current approved browser page; not a complete axe-core audit. |
| browser_performanceA | Read-only browser timing and slow resource summary of the current approved page. |
| context7_docsB | Opt-in read-only Context7 documentation lookup using a fixed HTTPS endpoint; never sends project files. |
| browser_axe_auditA | Run installed axe-core WCAG audit on the current locally approved browser page (read-only). |
| design_profilesB | Browse original design-inspiration profiles based on awesome-design-md. |
| design_readB | Read a project's DESIGN.md without modifying it. |
| design_initializeB | Create a project DESIGN.md from an inspiration profile, only if absent. Never overwrites. |
| design_auditC | Read-only scan for potential UI consistency issues; does not alter project files. |
| design_library_catalogC | Discover every current upstream Awesome DESIGN.md reference (refresh optional). |
| design_library_readB | Read a full version-pinned upstream DESIGN.md by catalog ID. |
| design_library_importA | Save an upstream design reference in .design-references without overwriting the approved DESIGN.md. |
| design_library_guidanceB | Produce a project-aware design implementation plan using the full reference and current project audit. No source edits. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 40 tools
Tools are grouped into clear sub-domains (Codex history, filesystem, browser, terminal/tasks, design, integrations) and names mostly specify both resource and action, so most selections are unambiguous. A few boundaries blur: design_profiles vs design_library_catalog and browser_snapshot vs browser_source_probe vs browser_diagnostics require reading descriptions to differentiate, and design_library_guidance overlaps conceptually with design_audit.
Dominant convention is snake_case verb_noun (read_file, patch_file, search_files, browser_click, design_initialize), with consistent prefixes per sub-domain (browser_, design_, design_library_, terminal_, project_task_). Minor deviations are noun-first names such as file_info, browser_status, codex_storage_info, and integration_catalog, which break the pattern without harming readability.
40 tools is heavy for a single server, and the surface spans at least eight loosely related domains (history indexing, filesystem, Playwright automation, command execution, design references, static analysis, docs). While each tool is non-redundant, the volume forces agents to sift a large catalog and suggests the server should be split by concern.
Each sub-domain has near-full lifecycle coverage: plan/run pairs for terminal and project tasks, scan/search/get/list for Codex history, and open/snapshot/click/fill/diagnostics/close for the browser, plus a coherent design-reference workflow (catalog, read, import, initialize, audit, guidance). Gaps exist—no delete/rename/move for files, no mkdir, and git is inspect-only—but these appear deliberate (allowlist safety) and are workable around.