@hiveai/mcp

hAIve MCP server - policy-aware briefing and memory tools for MCP-compatible coding-agent harnesses.

The MCP server is how agents load team policy before changing code. By default it exposes a small harness-oriented tool surface: briefing, relevant memories, failed-attempt capture, anchor verification, code-map lookup, and pre-commit checks. Larger maintenance and experimental surfaces are opt-in via HAIVE_TOOL_PROFILE.

hAIve is not just a memory database. The MCP layer participates in context policy: state-changing hAIve tools require get_briefing or mem_relevant_to first, so agents cannot silently skip team context while using hAIve.

Install

Recommended: install only @hiveai/cli. The MCP server is bundled inside haive — configure clients with command: "haive" and args: ["mcp", "--stdio"] (see @hiveai/cli README).

Standalone package (legacy / advanced):

npm install -g @hiveai/mcp

You usually still want the CLI for haive init, haive sync, etc.:

npm install -g @hiveai/cli

Quick start

# 1. Install the CLI
npm install -g @hiveai/cli

# 2. Initialize hAIve in your project (strict enforcement ON by default)
cd my-project
haive init          # .ai/, policy config, hooks, CI workflow, code-map
haive enforce install
# haive init --manual  # if you want to approve memories yourself

# 3. Point your AI client at the MCP server (see Client configuration below)

# 4. Bootstrap project context — run bootstrap_project prompt in your AI client once

# 5. Start every substantive task with get_briefing or mem_relevant_to

Client configuration

Claude Code

Add to ~/.claude.json (global) or .claude/settings.json (per-project):

{
  "mcpServers": {
    "haive": {
      "command": "haive",
      "args": ["mcp", "--stdio", "--root", "/absolute/path/to/your/project"]
    }
  }
}

Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "haive": {
      "command": "haive",
      "args": ["mcp", "--stdio", "--root", "/absolute/path/to/your/project"]
    }
  }
}

VS Code

code --add-mcp '{"name":"haive","command":"haive","args":["mcp","--stdio","--root","/absolute/path/to/project"]}'

Project-scoped (auto-detected)

Add a .mcp.json at the project root:

{
  "mcpServers": {
    "haive": {
      "command": "haive",
      "args": ["mcp", "--stdio"],
      "env": { "HAIVE_PROJECT_ROOT": "/absolute/path/to/your/project" }
    }
  }
}

The project root can also be set via the HAIVE_PROJECT_ROOT environment variable, or auto-detected from the nearest .ai/, .git/, or package.json.

Default MCP Tools

By default, hAIve runs with HAIVE_TOOL_PROFILE=enforcement. This keeps the agent surface small and aligned with the product promise.

Default tools:

• get_briefing
• mem_relevant_to
• mem_save
• mem_tried
• mem_search
• mem_get
• mem_verify
• code_map
• pre_commit_check
• mem_session_end

Default prompts:

• bootstrap_project
• post_task

Tool Profiles

HAIVE_TOOL_PROFILE controls how much hAIve surface an agent sees:

• enforcement (default): compact repo-native context harness for coding agents.
• maintenance: adds corpus review, lifecycle, distillation, code-search, and project-context maintenance tools.
• experimental: adds exploratory diagnostics such as runtime journal, pattern detection, why-this-file, why-this-decision, and conflict analysis.
• full: legacy alias for experimental.

Use maintenance for human/team stewardship sessions and experimental only when you are intentionally working on hAIve's broader research tooling.

get_briefing ⭐ Start every task with this

One-shot policy briefing: returns project context + module contexts + ranked decisions, gotchas, failed attempts, stale warnings, and setup warnings under a token budget. This is the first call agents should make before substantive edits.

{
  "task": "add a Stripe payment integration",
  "files": ["src/payments/PaymentService.ts"],
  "symbols": ["PaymentService", "TenantFilter"],
  "max_tokens": 8000,
  "max_memories": 8,
  "format": "full",
  "semantic": true
}

Parameters:

| Parameter | Default | Description | |---|---|---| | task | — | What you're about to do. Used to rank memories by relevance. | | files | [] | Files you're editing. Surfaces memories anchored to these files. | | symbols | [] | Symbol names to look up in the code-map (e.g. ["PaymentService"]). Returns file + line + kind without grepping. Requires haive index code. | | max_tokens | 8000 | Token budget for the entire response. Sections are truncated to fit. | | max_memories | 8 | Max memories to include. | | format | "full" | "full" = complete bodies · "compact" = 1-line summaries (call mem_get for details) | | semantic | true | Use embedding-based ranking if @hiveai/embeddings is indexed. | | include_stale | false | Include stale memories (may be outdated). | | track | true | Increment read_count for returned memories. |

Response includes:

• last_session — most recent haive session end recap (surfaced first so agents start with fresh context)
• project_context — .ai/project-context.md (suppressed if still template — is_template: true)
• module_contexts — relevant .ai/modules/<name>/context.md files
• memories — ranked memories with confidence, unverified flag (for draft/proposed), and match reason
• symbol_locations — file:line:kind results for each requested symbol (from code-map)
• decay_warnings — memory IDs not read in >90 days
• setup_warnings — actionable warnings (e.g. template project-context, missing init)
• search_mode — "semantic" | "literal_fallback" | "literal"

mem_save

Save a policy-relevant piece of knowledge. For failed approaches, use mem_tried immediately so the next agent sees the trap before repeating it.

Autopilot mode: memories go directly to validated with team scope by default. No approval cycle.

{
  "type": "gotcha",
  "slug": "open-in-view-false",
  "scope": "team",
  "body": "spring.jpa.open-in-view=false is intentional — do not re-enable. Lazy loading outside transactions causes N+1 queries.",
  "paths": ["src/main/resources/application.properties"],
  "tags": ["spring", "jpa", "performance"],
  "topic": "jpa-config"
}

| Parameter | Required | Description | |---|---|---| | type | ✅ | convention · decision · gotcha · architecture · glossary | | slug | ✅ | Short kebab-case identifier used in the file name | | scope | — | personal (default in manual mode) · team · module | | body | ✅ | Markdown content of the memory | | paths | — | Source file paths to anchor to — enables staleness detection by haive sync. Warning if path doesn't exist. | | symbols | — | Function/class names to anchor to | | tags | — | Tags for filtering and search | | topic | — | Stable key for upsert: if a memory with this topic+scope exists, update it in-place (revision_count++) | | domain | — | Business domain (e.g. payments) | | author | — | Author handle |

Response: { id, scope, file_path, action: "created"|"updated", warning?, invalid_paths? }

Deduplication: identical body content within the same scope is rejected. Use mem_update to modify an existing memory.

mem_tried ⭐ Record failures immediately

Record a failed approach. Automatically surfaces first in future get_briefing calls so agents don't repeat the same mistake.

{
  "what": "using require() to import gray-matter in an ESM package",
  "why_failed": "The package is ESM-only — require() throws ERR_REQUIRE_ESM",
  "instead": "Use import matter from 'gray-matter' (named default import)",
  "scope": "team",
  "paths": ["src/parser.ts"]
}

Auto-validated — no approval cycle needed.

mem_search

Search memories by substring or semantic similarity.

{
  "query": "flyway migration",
  "scope": "team",
  "semantic": true,
  "limit": 10
}

Falls back to literal search if embeddings are not indexed.

mem_get

Fetch a single memory with full body, anchor, confidence, and usage stats.

{ "id": "2025-01-15-gotcha-flyway-strict" }

mem_list

List memories with optional filters.

{
  "scope": "team",
  "type": "gotcha",
  "status": "validated",
  "tags": ["payments"]
}

mem_for_files

Given the files you're editing, return relevant memories grouped by reason (anchor overlap, module, domain).

{
  "files": ["src/payments/PaymentService.java", "src/payments/WaveProvider.java"]
}

mem_update

Update a memory's body, tags, or anchor without changing its id or usage history.

{
  "id": "2025-01-15-gotcha-flyway-strict",
  "body": "Updated explanation...",
  "paths": ["src/main/resources/db/migration"]
}

mem_verify

Check if anchor paths and symbols still exist in the current code. Detects stale memories and suggests possible renames when files have moved.

{ "id": "2025-01-15-gotcha-flyway-strict", "update": true }

mem_diff

Compare two memories side-by-side: shows frontmatter fields that differ and lines unique to each body. Useful before merging duplicates.

{ "id_a": "2025-01-15-gotcha-flyway-strict", "id_b": "2025-02-01-decision-flyway-naming" }

mem_session_end

Save a structured end-of-session recap. Topic-upsert: one recap per scope is kept and updated with revision_count++. Automatically surfaced at the top of the next get_briefing.

{
  "goal": "Add Stripe payment integration",
  "accomplished": "PaymentService done, tests passing, deployed to staging",
  "discoveries": "Webhook signature requires raw body, not parsed JSON",
  "files_touched": ["src/payments/PaymentService.ts", "src/payments/webhook.ts"],
  "next_steps": "Add retry logic for failed webhooks",
  "scope": "team"
}

mem_observe

Capture a code-level discovery in structured form (found-while, not a convention or decision).

{
  "file": "src/payments/PaymentService.ts",
  "symbol": "processPayment",
  "observation": "This method calls the external API synchronously — any timeout blocks the entire request thread.",
  "scope": "team"
}

mem_approve / mem_reject / mem_pending / mem_delete

Lifecycle operations:

{ "id": "2025-01-15-gotcha-flyway-strict" }               // mem_approve
{ "id": "2025-01-15-gotcha-old", "reason": "Outdated" }   // mem_reject
{}                                                          // mem_pending (list all)
{ "id": "2025-01-15-gotcha-old" }                          // mem_delete

get_project_context

Read .ai/project-context.md directly (without token budgeting).

{ "module": "payments" }   // Also loads .ai/modules/payments/context.md

bootstrap_project_save

Persist a project (or module) context document generated by the AI.

{
  "content": "# Project context\n\n## Architecture\n...",
  "module": "payments"   // Optional: save as .ai/modules/payments/context.md
}

code_map

Browse the pre-computed code map (file → exports + descriptions) instead of grepping.

{ "query": "payment" }              // Filter by keyword
{ "file": "src/payments" }          // Filter by file prefix
{ "symbol": "PaymentService" }      // Find a specific export

Requires haive index code to be run first.

MCP Prompts

post_task ⭐ Run before closing every session

Post-task reflection checklist. Guides the AI through capturing failed approaches, conventions, decisions, and gotchas before the session ends.

Use the post_task prompt with:
  task_summary: "Added Stripe payment integration"
  files_touched: ["src/payments/StripeService.ts", "src/payments/PaymentController.ts"]

bootstrap_project

Instructions for the AI to analyze the current project and save a structured context document to .ai/project-context.md. Run once after haive init.

import_docs

Analyze documentation (README, ADR, wiki page, API spec) and save actionable knowledge as hAIve memories.

Use the import_docs prompt with:
  content: "<full document text>"
  source: "docs/architecture.md"
  scope: "team"

The AI extracts up to 10 memories (conventions, decisions, gotchas, architecture) and calls mem_save for each.

Memory lifecycle

mem_save / mem_tried  → draft (personal) or proposed (team)
mem_approve           → validated
mem_verify            → stale (if anchors broken)
mem_reject            → rejected

Validated team memories appear in get_briefing. Stale memories are excluded by default (pass include_stale: true to override).

License

MIT