Mimiran exposes 38 structured tools over the Model Context Protocol. AI agents query routes, schemas, features, and invariants directly — no file parsing, no hallucinated context, no outdated snapshots.
A complete coding loop: query context → write code with correct data → sync changes back. All three steps happen over the same MCP connection.
The AI agent (Cursor, Windsurf, Claude Code, etc.) calls a Mimiran Read tool via the MCP protocol. No file system access required — context is returned as structured JSON.
→ CallTool context_by_route "/api/v1/payments"
← { route, handler, schema, feature, invariants }The agent uses the structured context to write code that respects actual table schemas, existing feature contracts, and architectural invariants — not hallucinated guesses.
→ CallTool db_describe_table "payments"
← { columns: [...], indexes: [...], enums: [...] }After writing code, the agent opens a delta-sync session and streams changed fragments back to Mimiran. Context stays accurate without a full rebuild.
→ CallTool open_sync_session {}
→ CallTool sync_session_batch { patches: [...] }
→ CallTool close_sync_session {}Query structured context about routes, features, and the full knowledge graph. These are the primary tools AI agents use during active coding sessions.
context_by_routeReturns full context for a given API route or page path — handler, schema, feature links, invariants.
context_by_featureReturns context scoped to a named feature — spec, related routes, DB tables, ownership.
search_contextSemantic search across the project knowledge graph. Returns ranked fragments by relevance.
get_feature_registryReturns the full FEATURE_REGISTRY.md as a structured object — all features, statuses, owners.
get_tech_stackReturns the complete technology stack snapshot — runtimes, frameworks, infrastructure.
get_ai_rulesReturns AI coding rules and constraints defined for this project. Agents respect these by default.
get_ui_rulesReturns UI and design system rules — component conventions, token usage, layout constraints.
Expose the full database schema to AI agents without sending migration files. Tables, columns, types, and enums — structured and queryable.
db_list_tablesLists all tracked database tables with their column counts and last-synced timestamps.
db_describe_tableReturns full schema for a given table — columns, types, nullability, foreign keys, indexes.
db_list_enumsLists all tracked database enums with their values and the tables that reference them.
Query the architectural rules defined for your project. Invariants are validated by the CLI and exposed to agents so they can self-check before writing code.
get_invariantsReturns all defined architectural invariants — forbidden patterns, import rules, platform isolation constraints.
get_context_metaReturns context health metadata — last sync timestamps, artifact freshness, coverage gaps.
get_setup_stateReturns the current setup wizard state — completed steps, missing configuration, next actions.
Inspect project binding, connected repositories, and snapshot availability. Used during setup and by orchestration agents.
get_project_bindingReturns the active project binding — project ID, org, server endpoint, token scope.
list_repo_connectionsLists all registered repository connections for the current project.
list_snapshotsLists all available context snapshots — version, created timestamp, artifact coverage.
get_snapshot_statusReturns the build status of a specific snapshot — processing, ready, or failed.
get_sync_helper_stateReturns the sync helper adapter installation state and version on the host.
get_sync_session_stateReturns the state of the current active sync session — open batches, pending flushes.
get_pending_bridge_uploadReturns any pending bridge upload task — token, expiry, expected artifacts.
Trigger full sync jobs that rebuild artifacts from the codebase. These are the batch-mode counterparts to delta-sync sessions.
sync_monorepo_mapRebuilds MONOREPO_MAP.md from the current repo structure — packages, apps, shared modules.
sync_contextSynchronises context/* with codebase — feature specs, page docs, dependency graph, DB schema.
start_sync_jobEnqueues a named sync job on the server for async execution.
execute_sync_jobExecutes a specific sync job synchronously and returns the result.
Incremental sync for active AI coding flows. Instead of rebuilding all artifacts, sessions stream only the changed fragments — keeping context fresh without full scans.
open_sync_sessionOpens a new incremental sync session. Returns a session ID used by subsequent batch calls.
sync_session_batchSends a batch of changed fragments within an open session. Supports partial file updates.
flush_sync_sessionFlushes buffered session data to the server without closing the session.
close_sync_sessionCommits all session changes and closes the session. Triggers artifact reindex.
Generate structured markdown scaffolds that agents fill in. These tools create the skeleton — agents populate the content based on current code context.
scaffold_feature_specCreates a feature spec scaffold in context/features/ with sections for purpose, routes, DB tables, ownership.
scaffold_page_docCreates a page documentation scaffold in context/pages/ with sections for layout, data flow, components.
Bootstrap and configure projects programmatically. Used by setup agents and CI/CD pipelines to register repositories and initialise context infrastructure.
setup_projectRuns the full guided project setup — creates project, registers repo, uploads initial snapshot.
create_projectCreates a new project in the current org with the given name and settings.
delete_projectDeletes a project and all its associated artifacts, tokens, and sync state.
connect_projectConnects the local repo to an existing project by binding the project ID.
upsert_repo_connectionRegisters or updates a repository connection — remote URL, branch, root path.
install_sync_helperInstalls the sync helper adapter on the host machine for the current project.
ensure_host_adapterVerifies the host adapter is installed and at the correct version, installs if missing.
write_bridge_tokenWrites a one-time bridge upload token to disk for the initial source snapshot transfer.
Every MCP call goes through Mimiran's security layer before reaching project data.
ReadResource is protected against path traversal attacks. Files are served only from allowed directories, never outside project context.
All MCP calls are logged with secret masking and prompt/query truncation. Sensitive data never appears in plain logs.
Readonly mode blocks all Write tools, enabling safe audit runs without any risk of accidental state mutations.
Each MCP token is scoped to a single project. Tokens cannot cross project boundaries — no lateral access between repos.
Connect Mimiran to your project and every MCP tool is immediately available in Cursor, Windsurf, Claude Code, and any other MCP-compatible agent.