NAVIG
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| NAVIG_DATA_DIR | No | Override the data directory (default is ~/.navig/). | |
| NAVIG_CONFIG_DIR | No | Override the global config directory (default is ~/.navig/), which holds config.yaml, vault, sessions, workspace, and logs. | |
| NAVIG_NO_TELEMETRY | No | Disable the one-time anonymous install ping. | 1 |
| NAVIG_SKIP_ONBOARDING | No | Skip the first-run setup wizard. Required for non-interactive starts. | 1 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
| resources | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| navig_fleetA | Inspect the infrastructure this NAVIG instance is configured for: SSH hosts, the apps behind them, recent errors/context and database connections. Read-only. Use navig_knowledge for documentation and navig_runtime for missions; navig_run_command is only exposed by the full catalogue. Returns host/app/database entries (names, addresses, paths, status); pass a name to host_info/app_info. |
| navig_knowledgeA | Search and read NAVIG's local knowledge: product documentation, wiki pages, and the linked files a Space exposes. Read-only. Use navig_recall for what was said or decided earlier, navig_fleet for infrastructure inventory. Returns matching passages/pages with their source path; pass a path to read_wiki_page, or a Space id to the space_* actions. |
| navig_recallA | Search the owner's recorded past: conversations, decisions, tables, open questions, forgotten ideas and source excerpts, every result cited. Read-only. Use navig_knowledge for docs/wiki and navig_memory for durable facts. Returns records with ids and citations; pass an id to get/related/timeline or a citation to source_excerpt. |
| navig_memoryA | Read NAVIG's durable key-fact memory: matching facts with their ids/scores, or counts. Read-only. Use navig_memory_update to create/change/retire a fact, and navig_recall for conversation history. Returns matching facts with ids and scores, or {total, active, deleted, superseded} counts. |
| navig_runtimeA | Read NAVIG runtime state: registered Nodes, Missions, ExecutionReceipts and a Node's TrustScore. Read-only. Use navig_runtime_action to create or advance a Mission. Returns nodes/missions/receipts with ids and state; pass a node_id to trust_score or a mission id to receipt listing. |
| navig_automationA | Browse the NAVIG Harbor Bay catalogue and installed Blocks, including a block's typed inputs, ordered steps and verify criterion. Read-only. Use navig_automation_apply to acquire an item or run a block. Returns items/specs with slugs, ids, risk and install state. |
| navig_webA | Search and fetch the public web, or run Firecrawl search/scrape/crawl. Outbound network access; Firecrawl usage is billable and crawl starts a short-lived job. Use navig_connectors for authenticated provider data and navig_knowledge for local docs. Returns ranked results, page content/metadata, or crawl job pages. |
| navig_browser_debugA | Inspect a NAVIG-launched Chrome/Edge/Brave/Electron debug browser: live targets, open tabs, accessibility snapshots with refs, and which browsers are running. Read-only. Use navig_browser_debug_act to interact with that browser, and navig_browser_user for the operator's own browser. Returns ports/tabs/tree entries; pass a ref to the act router. |
| navig_browser_userA | Work with the operator's own browser through Dock: list tabs, read snapshots/text, read console/network (the first call starts capture), screenshot, or act (open/click/type/scroll) under a visible overlay. Screenshots write a file. Use navig_browser_debug for a NAVIG-launched debug browser; this never silently falls back to one. Returns tab ids, page text/tree, capture entries, or per-step results. |
| navig_connectorsA | Read from connected provider accounts (Gmail, Drive, Slack, GitHub, Notion, Linear, Supabase, Calendar, …). Read-only; capabilities vary by installed connector. Use navig_connectors_act to change provider data. Returns provider status, or matching items with native ids usable by fetch and act. |
| navig_memory_updateA | Create, update or soft-delete a durable key fact in NAVIG memory. State-changing and audited. Use navig_memory to read facts first. Returns the stored fact id and status; forgetting is a soft delete. |
| navig_runtime_actionA | Create a Mission or advance its state (start/succeed/fail/cancel/timeout). State-changing; mission_action requires approval. Use navig_runtime to read missions and receipts first. Returns the Mission with its new state. |
| navig_automation_applyA | Acquire a Harbor Bay item or apply an installed Block end-to-end. State-changing: may install content or run steps; a Block needs its step approvals and resolves secrets only from env/vault. Use navig_automation to inspect first. Returns the execution receipt or the acquisition action. |
| navig_browser_debug_actA | Act in a NAVIG-launched debug browser: launch/new/navigate/click/type/key/scroll/move/screenshot/switch, detach/stop, or evaluate/inject/login. State-changing and high-risk; injected scripts persist until removed and login fills a real credential (never returned). Use navig_browser_debug to inspect and get refs first. Returns the dispatch result; re-snapshot to verify. |
| navig_connectors_actA | Change data in a connected provider account: reply/create/update/delete/archive/label/send/move, whichever the provider supports. State-changing and dangerous; approval required. Use navig_connectors to find the resource id first. Returns success plus the affected resource, or success:false with the reason. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| NAVIG Hosts Configuration | All configured SSH hosts |
| NAVIG Apps Configuration | All configured applications |
| NAVIG Wiki | Project knowledge base |
| NAVIG Context | Current system state and recent errors |
| NAVIG Agent Status | Agent runtime status and configuration |
| NAVIG Agent Goals | Autonomous goal list and progress |
| NAVIG Agent Remediation | Remediation actions and recent remediation logs |
| NAVIG Agent Learning Report | Latest error pattern analysis report |
| NAVIG Agent Service Status | Service installer/status integration |
| Agent Status | Alias for NAVIG agent runtime status |
| Agent Goals | Alias for agent goals |
| Agent Remediation | Alias for remediation actions |
| Agent Learning Patterns | Alias for learning report |
| Agent Service | Alias for service status |
| NAVIG Runtime Nodes | All registered Node identities |
| NAVIG Runtime Missions | Recent Missions with lifecycle state |
| NAVIG Execution Receipts | Audit trail of completed Mission executions |
TDQS
Scored across 15 tools
Each tool targets a distinct domain+action, and descriptions explicitly cross-reference neighbors to prevent confusion (e.g. navig_memory vs navig_recall vs navig_knowledge, navig_browser_debug vs navig_browser_user vs navig_web). The read/act pairing (debug vs debug_act, connectors vs connectors_act, runtime vs runtime_action) is clearly delineated. Boundaries are stated rather than left implicit.
Reads follow a uniform navig_<domain> pattern (navig_fleet, navig_knowledge, navig_recall, navig_web). Writes deviate slightly in verb suffix: navig_connectors_act and navig_browser_debug_act use '_act', navig_runtime_action uses '_action', while navig_memory_update and navig_automation_apply use other verbs. Mostly consistent with minor deviations.
15 tools sit at the top of the well-scoped range, and each earns its place: nine distinct read domains plus their corresponding state-changing counterparts. The read/act split is principled rather than redundant. No filler tools.
Strong lifecycle coverage: memory (read/update), runtime (read/action), automation (browse/apply), connectors (read/act), browser (inspect/act). Minor gaps remain—knowledge, recall, fleet and web are read-only with no write path, and memory lacks an explicit hard-delete. These are largely by design and workable.