x-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| X_MCP_MODE | No | Optional mode for the MCP server. `auto` (default) uses CDP only when X_MCP_CDP_URL is set; otherwise it uses the harness handoff. `browser` forces direct browser mode using a local Chrome DevTools Protocol endpoint. `harness` forces harness handoff. | auto |
| X_MCP_CDP_URL | No | Optional local Chrome DevTools Protocol (CDP) endpoint URL for direct browser mode, e.g. http://127.0.0.1:9222. Only explicit loopback endpoints are accepted. |
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
} |
| prompts | {
"listChanged": true
} |
| resources | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| x_search_topicA | Start here for X/Twitter research. In harness mode returns needs_browser with optimized Latest URLs and extraction instructions: then use YOUR browser and call x_collect_posts. With configured CDP returns observed posts. Never treat a plan as tweets. |
| x_collect_postsA | After x_search_topic and browser reading, submit exact captured posts. Merges snapshots in memory, removes duplicates, rejects stale/unrelated posts, returns dates and source links. Captures must use the plan’s search URLs. Does not fetch or independently authenticate host-supplied text. |
| x_get_statusA | Reports configuration and browser requirements; does not access X or test login. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| research_topic | Research a topic using recent, cited X posts and the host browser. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| workflow | How the host uses its own browser to collect recent X posts. |
| extractor | Read-only rendered-page extraction script for a harness browser evaluator. |
TDQS
Scored across 3 tools
The three tools occupy distinct roles in a pipeline: plan generation (x_search_topic), post ingestion (x_collect_posts), and config reporting (x_get_status). Boundaries are clear from descriptions, though the 'search' name on x_search_topic is slightly misleading since it returns a plan rather than results, which could momentarily confuse an agent about where actual data comes from.
All tools use a uniform x_verb_noun pattern (x_search_topic, x_collect_posts, x_get_status), with consistent prefixing and snake_case throughout. There are no deviations or mixed conventions.
Three tools is compact but matches the narrow, deliberately minimal harness workflow of plan→capture→status. Each tool earns its place, though the surface is on the lean side with no room for auxiliary operations.
The plan→collect→status lifecycle is closed, but the actual fetching is delegated entirely to the agent's external browser, so the server cannot complete a research task on its own. There is no post detail retrieval, refresh/clear operation, or recovery path if collection fails, leaving notable gaps for the stated research purpose.