apogee-mcp

MCP server for AI coding assistants, automated review agents, and CI pipelines. Exposes 27 tools, 8 resources, and 2 prompts over stdio.

MCP surface: This is the MCP server — all Apogee MCP tools originate here.

Install: pip install apogee (included with the Apogee server install)

apogee-mcp

Synopsis

apogee-mcp [--transport stdio]

Start the server. Communicates via stdio transport with any MCP-compatible host (Claude Code, Cursor, OpenCode).

Description

The MCP server is the integration surface for three use cases:

1. AI-assisted development — coding assistants (Claude Code, Cursor) use skills and chains for structured workflows with human-in-the-loop approval.

2. Automated code review — review agents call CPG tools (cpg_full_audit, cpg_backport_check, cpg_security_scan) to produce structured findings on PRs with zero human prompting.

3. Continuous audit — CI pipelines index commits incrementally and run security scans against baselines, flagging regressions automatically.

It provides 27 tools, 8 resources, and 2 prompts.

Constraint system: the route tool classifies each user message as read or write intent. In read mode, only discovery and inspection tools are advised. In write mode, the full skill catalog is available. Constraint transitions are automatic (auto mode) or explicit (force-write).

Configuration

The server reads APOGEE_ROOT from the environment to locate the project root. When not set, it uses the current working directory.

MCP client configuration example:

{
  "mcpServers": {
    "apogee": {
      "command": "apogee-mcp",
      "args": [],
      "env": { "APOGEE_ROOT": "/path/to/project" }
    }
  }
}

Reference

Workflow tools

Tool	Description	Key parameters
route	Classify user intent and update session constraint	message
find_skill	Search skills by natural language query	query, top_k
run_skill	Primary execution entrypoint — prepare and run a skill	workflow_id
suggest_next	Recommend next skill based on project state	(none)
set_constraint	Set session constraint explicitly	constraint, mode

Chain tools

Tool	Description	Key parameters
define_chain	Derive lifecycle chain from plans	chain_type, feature_slug
execute_chain	Manage live chain session	action, chain_handle
checkpoint_chain	Persist chain session for later resumption	session_id
resume_chain	Restore saved chain checkpoint	state_path

Manifest tools

Tool	Description	Key parameters
generate_manifest	Create repo structural analysis manifest	repo, label, out_path
rescore_manifest	Recompute manifest scores without regenerating	manifest_path
summarize_manifest	Return score summary from manifest	manifest_path
get_manifest_context	Load budget-aware manifest fragments	label, budget_tokens

CPG tools

Tool	Description	Key parameters
cpg_find_callers	Find functions calling a given function	function_name, depth, branch
cpg_find_callees	Find functions called by a given function	function_name, depth, branch
cpg_impact_analysis	Analyze change blast radius	function_name, file, branch
cpg_show_module_structure	Show file layout: modules, classes, functions	path, branch
cpg_trace_data_flow	Trace variable flow within and across functions (follows arg_passes_to edges)	variable, function, file, branch
cpg_temporal_diff	Compare CPG state between commits	commit_a, commit_b, file_path, subsystem, branch
cpg_semantic_search	Find structurally similar functions	query_function, k, subsystem, branch
cpg_reconstruct_at_commit	Reconstruct file/function body at a historical commit	commit_hash, file_path, function_name
cpg_raw_query	Execute read-only SQL against CPG database	sql, params, max_rows
cpg_security_scan	Run compound security scan using scope_summary flags	subsystem, branch, min_risk
cpg_branch_divergence_report	Cross-branch structural divergence report	branch_a, branch_b, subsystem
cpg_backport_check	Assess backport safety for a function	function_name, file_path, source_branch, target_branch
cpg_full_audit	Combined security + divergence + backport audit	branch_a, branch_b, report_level
cpg_function_complexity	Rank functions by structural complexity	file_path, subsystem, branch, min_complexity

Resources

URI	Description
apogee://workflows	Compiled workflow catalog by collection
apogee://workflows/{id}	Metadata and prompt text for one workflow
apogee://status/repo	Lifecycle dashboard: features, manifests, chains, gates
apogee://status/server	Server health, registry size, source layers
apogee://status/gates	Acceptance gate evaluation results
apogee://status/constraint	Session constraint state
apogee://status/session	Session constraint state and active persona
apogee://status/config-health	IDE config hygiene scan

Prompts

Prompt	Description
prepare_workflow	Prepare workflow for execution — returns rendered prompt
run_workflow	Prepare and render workflow prompt text

Examples

# Discover a skill by intent
find_skill("commit feature")
# Returns: workflow_id="task.git.commit_feature", score=1.0

# Execute a skill
run_skill("task.git.commit_feature")
# Returns: prompt_text, resolved_inputs, expected_artifacts

# Route a message (call on every user message)
route("refactor the parser module")
# Returns: intent="write", constraint="write", matched_skill=...

Note

This page is also available as a man page: man apogee-mcp

See Also

• Chain Engine — lifecycle chain execution

• Skill Runtime — skill compilation and registry

• apogee-manifest — manifest generation tools

• CPG Engine — CPG query tools