Context Bridge
<!-- mcp-name: io.github.jusspound-netizen/context-bridge -->
# Context Bridge
**Context Bridge - one MCP endpoint for all your dev tools**
**Problem:**
* Tool sprawl across Jira, Sentry, GitHub, and CI systems.
* Context fragmentation makes it hard for AI agents to operate across tools.
* AI agents are currently blind to critical development context (Jira tickets, Sentry alerts, CI status).
**Setup:**
1. `pip install -r requirements.txt` (or necessary dependencies)
2. Set required environment variables (e.g., API keys for Jira/Sentry).
3. Add the running MCP server endpoint to your AI environment (e.g., Claude Code or Cursor) as a custom MCP server.
## Quick Start
To run the server:
1. `python server.py`
2. Ensure required environment variables are set.
## Example
A sample response from `get_jira_ticket`:
{"id": "PROJ-123", "summary": "Fix login bug", "status": "In Progress"}
**Tools:**
| Tool Name | Description | Status |
| :--- | :--- | :--- |
| `get_jira_ticket` | Retrieves mock Jira ticket data by ID. | ✅ Implemented |
| `get_sentry_issues` | Retrieves Sentry alerts. | ✅ Implemented |
| `get_github_actions_runs` | Checks CI/CD status for a repo. | ✅ Implemented |
| `send_slack_notification` | Sends status updates to Slack. | ✅ Implemented |
**License:** MIT
TDQS
Scored across 3 tools
Each of the three tools targets a different external service (Jira, GitHub Actions, Sentry) with an unambiguous, non-overlapping purpose. There is no realistic scenario in which an agent would confuse retrieving a Jira ticket with fetching CI runs or Sentry issues.
All three tools follow the same get_<source>_<resource> snake_case pattern (get_jira_ticket, get_github_actions_runs, get_sentry_issues). The convention is predictable and self-describing, so an agent can guess a tool name from its intent.
Three tools is at the thin end for a server framing itself as a 'context bridge' — each tool earns its place, but the surface is narrow. A single tool per integration means there is no fallback or alternative granularity, making it borderline under-scoped.
Read-only retrieval across three sources is coherent, but the surface has notable gaps for the stated purpose: only Actions runs (no PRs, commits, or issues), only unresolved Sentry issues (no detail or resolution filtering), and no pagination/filter parameters documented. Agents assembling cross-source context will hit dead ends quickly.