Skip to content

Tool Reference

nspec exposes 92 MCP tools. Every tool below is generated from the live @mcp.tool registry in src/nspec/interfaces/mcp/*.py, so this page stays in lockstep with the server.

Key args list each tool's parameters; a trailing * marks a required argument. Universal plumbing args (e.g. docs_root) are omitted.

Regenerate after adding tools: the parity drift test (S591) fails CI if this page and the registry diverge.

Query & Discovery

Tool Purpose Key args
epics List all epics with progress tracking.
epic_specs Query specs belonging to an epic, optionally filtered by status. epic_id*, status_filter, starting_spec_id
show Show full details for a spec (FR + IMPL). spec_id*
next_spec Find the next spec to work on based on priority and blockers. epic_id
get_epic Get the currently active epic from the active session.
validate Run the 6-layer validation engine on all specs.
estimate_loe Estimate Level of Effort for a spec. spec_id*
blocked_specs List specs that have blocked tasks, are paused, or in exception state.
dep_graph Analyze dependency graph across active specs. epic_id, spec_id
preflight Pre-flight state snapshot for a spec before /ngo execution. spec_id*
digest Run the nightly digest: collect activity, classify coverage, auto-file review specs. since, dry_run
pr_gate Run the PR completeness gate. pr_number, base_branch
loop_pipeline Return the resolved loop pipeline and agent map.
discover_providers Discover available provider models and configuration values. provider

Spec Lifecycle

Tool Purpose Key args
create_spec Create a new FR+IMPL spec pair. title*, priority, epic_id, is_epic, set_default, spec_type, template
activate Activate a spec by advancing its status and writing the active session. spec_id*
advance Advance a spec to the next logical status. spec_id*
complete Complete a spec and archive it to completed/done. spec_id*
supersede Supersede a spec and archive it to completed/superseded. spec_id, superseded_by
park Park a spec by setting IMPL status to Paused. spec_id*, reason
park_epic Park all non-completed specs in an epic. epic_id*, reason
exception Set spec to Exception state for human triage. spec_id, reason
hold Put a spec on hold by setting IMPL to Hold state. spec_id, reason
recover Recover a spec from Exception to Active state. spec_id*
reset Reset a spec back to Planning state. spec_id*
resume Resume a Paused spec to the status it held before it was parked. spec_id*
start_design Move FR to In Design without starting IMPL. spec_id*
force_archive Force-archive a spec that's stuck in an inconsistent state. spec_id, reason
move_spec Reassign a spec's owning epic, or detach it from any epic. spec_id*, target_epic_id
rename_spec Retitle a spec or epic — ID-stable heading + slug rename for both paired files. spec_id, new_title
delete_spec Hard-delete a spec (FR+IMPL) or an epic (FR-only) — the guarded escape hatch. spec_id*, confirm
unify_collisions Detect and resolve spec ID collisions.
init Initialize a new nspec project with full directory structure. epic_title, force

Tasks & Acceptance Criteria

Tool Purpose Key args
task_complete Mark an IMPL task as complete. spec_id, task_id, marker, run_tests
task_block Mark an IMPL task as blocked with a reason. spec_id, task_id, reason*
task_unblock Remove the BLOCKED marker from an IMPL task. spec_id, task_id
criteria_complete Mark an acceptance criterion as complete in the FR file. spec_id, criteria_id, marker, run_tests
set_impl_task Add, edit, or remove an IMPL task line — text CRUD only (no reorder). spec_id, action, phase, task_id, text, file, acknowledge_completed
set_acceptance_criteria Add, edit, or remove an acceptance-criterion line in an FR — text CRUD only. spec_id, action, section, label, text
spec_test_files Manage test file entries in the ### Test Files section of an IMPL. spec_id*, action, files
spec_tests Run only the tests relevant to a spec. spec_id*

Dependencies & Priority

Tool Purpose Key args
add_dep Add a dependency to a spec. spec_id, dep_id
remove_dep Remove a dependency from a spec. spec_id, dep_id
set_priority Set the priority (roadmap position) of a spec or epic. spec_id, priority

FR/IMPL Authoring

Tool Purpose Key args
refine_fr Check if an FR is a skeleton and generate an authoring prompt. spec_id*
refine_impl Check if an IMPL is a skeleton and generate a refinement prompt. spec_id*
write_fr_refinement Write agent FR authoring output to the FR file. spec_id, agent_output
write_refinement Write agent refinement output to the IMPL file. spec_id, agent_output
cheap_docs Generate documentation for a spec using a low-reasoning agent. spec_id*

Review

Tool Purpose Key args
review Execute an atomic review cycle for a spec. spec_id*, base_branch, test_results, implementer
review_spec Generate a review prompt for a spec to use with execute_agent(). spec_id*, base_branch, test_results
write_review_verdict Write a review verdict to the IMPL file. spec_id, verdict, feedback, reviewer, implementer, review_token, verdict_token

Sessions

Tool Purpose Key args
session_start Initialize a work session for a spec. spec_id*
session_save Persist session state for cross-session handoff. spec_id*, task_id, notes, phase
session_resume Load persisted session state for a spec. spec_id*
session_clear Archive the active session and remove the active link.
session_list List all sessions, optionally filtered by spec ID. spec_id
session_get Get a session by its session ID. session_id*
session_delete Delete a session file. session_id*
loop_init Initialize a loop session targeting a specific spec. spec_id*

Checkout (lease-based reservation)

Tool Purpose Key args
checkout Reserve a spec with a TTL lease (checkout without activating). spec_id*, agent_id, ttl
checkin Release a spec checkout reservation. spec_id*
checkout_heartbeat Extend the lease for a checked-out spec. spec_id*, ttl
checkout_status List all active spec checkouts with TTL remaining.

Agent Queue

Tool Purpose Key args
queue_init Initialize the agent assignment queue from the backlog. epic_id, max_agents, lease_ttl
queue_claim Claim the next unclaimed spec from the queue. agent_id*
queue_release Release a claimed spec back to the queue or mark it completed. agent_id, spec_id, reason, completed
queue_heartbeat Extend the lease for an agent's current claim. agent_id*
queue_status Get full queue status: agents, queued specs, completed specs.

Agent Swarm

Tool Purpose Key args
swarm_launch Launch a parallel Claude Code agent swarm (S253). epic_id*, num_agents, model
swarm_status Get the current swarm status: agents, progress, and health.
swarm_teardown Tear down the active swarm: kill agents, clean worktrees, clear state.

Agent Execution

Tool Purpose Key args
execute_agent Execute a prompt via the configured agent for the given purpose. prompt, prompt_file, timeout, model, reasoning_effort, sandbox, approval_policy, flags, purpose, review_token, spec_id

Skills

Tool Purpose Key args
skills_install Install skills to .novabuilt.dev/nspec/commands/ and sync to agents. force
skills_list List available skills with source info.
skills_sync Re-sync skills from sources (update after nspec upgrade). force
skills_update Apply skills update with user's consent choice. choice*

GitHub Issues

Tool Purpose Key args
issue_import Import a GitHub issue as an nspec spec. repo, number, epic_id, priority, dry_run
issue_list List GitHub issues from a repository. repo, state, label, limit
issue_sync Audit GitHub issue sync for an epic — find mismatches. epic_id*, repo
batch_import Batch-import all open non-imported GitHub issues as nspec specs. epic_id, repo, label
create_issue Create a GitHub issue from a spec and link it. spec_id*, repo

Artifacts

Tool Purpose Key args
artifacts_list List artifacts — per-spec work artifacts (when spec_id given) or nspec-managed templates. spec_id, artifact_type
artifacts_sync Sync artifacts — recover orphaned files or sync nspec templates. dry_run, spec_id

Audit & Quality

Tool Purpose Key args
audit_record Record an audit artifact (decision, inventory, or handoff). kind, topic, content*, related_specs
coverage_audit Run full-suite coverage audit and return candidate status. coverage_json, registry, strict, include, exclude
context_audit Analyze a CLAUDE.md file for context budget and extraction candidates. file, threshold
branch_status List all nspec/* branches with spec ID, PR status, and ahead/behind main.
branch_cleanup Remove nspec/* branches whose PRs have been merged. dry_run