Skip to content
Agentmart
Submit a tool

Context First MCP

37-tool MCP server: context health, memory, reasoning, truthfulness & research pipeline.

Connector by XJTLUmedia · GitHub account verified by the MCP Registry

IndexedMITLocal (stdio)★ 2
Write a review

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/xjtlumedia-context-first-mcp

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

37-tool MCP server: context health, memory, reasoning, truthfulness & research pipeline.

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list and install config. Checked Oct 25, 2026.

Tools
4 read · 0 write · 0 delete · 4 not yet classified
Secrets
None asked for
Install scripts
Not scanned yet
Runs as
Node.js package

Tools by what they do 8

Read from the source code; the running server may list different tools.

Read 4 · Look things up without changing anything.

  • context_health
  • export_research_files
  • reason
  • research_pipeline

Not yet classified 4 · We could not tell from the name and description.

  • context_loop
  • memory
  • sandbox
  • truthcheck

Tools (8)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

  • context_health

    [CONTEXT & STATE] 13 sub-tools: recap, conflict, ambiguity, verify, entropy, abstention, grounding, drift, depth, get_state, set_state, clear_state, history. Auto-selects based on params or use 'check' to override. TIP: context_loop runs all health checks automatically — prefer context_loop for comprehensive analysis, use context_health for targeted checks.

  • context_loop

    [ORCHESTRATOR — CALL THIS FIRST] CALL THIS TOOL every 2-3 turns and at the start of ANY task. It is the single most important tool — it replaces calling recap, conflict, ambiguity, entropy, grounding, drift, depth, and discovery tools individually. What it does: Runs ALL context health checks in one call. Auto-extracts facts from conversation, detects contradictions, checks answer quality, and tells you exactly what to do next. Returns a 'directive' object with: - action: proceed | clarify | reset | abstain | deepen | verify - instruction: Plain English telling you what to do - contextHealth…

  • export_research_files

    [EXPORT] Automatically writes research artifacts to disk. It can expand and write every verified report chunk without asking the LLM to loop finalize manually, and it can also write every gathered raw-evidence batch even when verify has not passed yet.

  • memory

    [MEMORY] 6 sub-tools: store (hierarchical ingest), recall (adaptive gate retrieval), compact (compress with integrity), graph (knowledge graph with PageRank), inspect (tier status), curate (importance-based curation). Auto-selects based on params or use 'action' to override. TOOL NAME: memory (use underscores).

  • reason

    [REASONING] 5 engines: inftythink (iterative bounded reasoning), coconut (multi-perspective latent analysis), extracot (reasoning chain compression), mindevolution (evolutionary search), kagthinker (structured logical decomposition with dependency DAG). Auto-selects based on params or use 'method' to override.

  • research_pipeline

    [PIPELINE] RECOMMENDED for research tasks. Orchestrates all underlying Context-First layers through 6 phases (init→gather→review→analyze→verify→finalize). NEW: plan→draft→review→fix loop — like compile→test→fix in coding. Init generates a research outline (12+ sections). Each gather adds depth to one section with quality gate (25K char / 500 line min — multiple gathers per section expected). Review runs quality tests and identifies gaps. CRITICAL: Interleave web search and gather — after EACH search, IMMEDIATELY call gather with deeply written content. Do NOT batch searches. Each gather writes…

  • sandbox

    [SANDBOX] 3 sub-tools: discover (semantic tool search via TF-IDF), quarantine (isolated state sandbox), merge (merge/discard silo). Auto-selects based on params or use 'action' to override.

  • truthcheck

    [TRUTHFULNESS] 7 tools: probe (linguistic truth signals), truth_direction (truth vector projection), ncb (perturbation robustness), logic (formal logical consistency), verify_first (5-dimension verification), ioe (confidence-based correction), self_critique (iterative refinement). Auto-selects or use 'check' to override. Set cascade=true for auto-correction on low scores.

Signals

Stars
2
Forks
0
Open issues
0
Last commit
Apr 15, 2026 (6 months ago)
Last release
Apr 14, 2026 (6 months ago)

From GitHub (fetched Oct 24, 2026).

Details

Registry name
io.github.XJTLUmedia/context-first-mcp
Version
1.2.2
Auth
No auth
Packages
npm context-first-mcp
License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from XJTLUmedia

Cite this page

Context First MCP. Agent Mart. https://agentmart.dev/connectors/xjtlumedia-context-first-mcp (last updated 2026-10-24)