codapult-guard
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| GUARD_APPROVER | No | Declared approver identity. When `requireDistinctActor` is enabled in `rules.json` (`approval.requireDistinctActor: true`), CLI and permitted MCP approvals require `GUARD_APPROVER`, and Guard rejects the same declared actor used to generate a proposal via `GUARD_PROPOSER`. These variables provide declared provenance only; external branch protection or signed identity remains responsible for proving who approved the change. | |
| GUARD_PROPOSER | No | Declared proposer identity, recorded as provenance when a proposal is generated. Used together with `GUARD_APPROVER` under `approval.requireDistinctActor` to reject an approval coming from the same declared actor that generated the proposal. These variables provide declared provenance only; external branch protection or signed identity remains responsible for proving who approved the change. |
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 | {
"listChanged": true
} |
| prompts | {
"listChanged": true
} |
| resources | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| codapult_guard_analyzeB | Refresh persisted project facts and snapshots without changing rules, contracts, or baseline. |
| codapult_guard_next_actionC | Return the next bounded action for an AI coding-agent loop without modifying project files. |
| codapult_guard_contextA | Read the discovered project model, architecture insights, and local rules used by Codapult Guard. Works before init by discovering the project on demand. |
| codapult_guard_proposeB | Generate evidence-based rule and contract proposals for an AI agent. Proposals are never activated automatically. |
| codapult_guard_initB | Create Guard artifacts after explicit confirmation; existing baselines are protected by default. |
| codapult_guard_proposal_decideB | Approve or reject current evidence-backed proposals; stale proposals cannot be activated. |
| codapult_guard_verifyC | Run project checks, production build, and architecture regression checks as one verification gate. |
| codapult_guard_governanceB | Audit declared proposal and approval provenance, policy bindings, and retained verification reachability. This reports metadata; it does not prove identity or signatures. |
| codapult_guard_auditA | Run a full Guard audit, ignoring baseline suppression and validating contracts. |
| codapult_guard_reviewC | Prepare a bounded semantic-review packet containing the current diff, project model, contracts, and deterministic findings for an AI reviewer. |
| codapult_guard_checkC | Check project architecture rules and return structured new violations. |
| codapult_guard_waiversC | List and manage time-bounded, project-approved exceptions to active Guard findings. |
| codapult_guard_impactC | Explain the project impact of one or more changed files using the discovered dependency graph, routes, capabilities, and contracts. |
| codapult_guard_explainC | Explain a Guard rule or contract with its evidence and applicable scope. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| codapult_guard_proposal_review | Refine evidence-backed Guard rule and contract proposals before activation. |
| codapult_guard_semantic_review | Review a Guard packet against project-specific contracts and architecture context. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| codapult_guard_runs | Local verification outcomes, gates, and stage timings. |
| codapult_guard_rules | Local architecture invariants available to CLI checks and AI agents |
| codapult_guard_context | Current project model, architecture policy, contracts, and gate settings. |
| codapult_guard_architecture | Observed architecture and conventions. |
| codapult_guard_waivers | Active and expired time-bounded exceptions to Guard findings. |
| codapult_guard_proposals | Evidence-backed rules and contracts awaiting review. |
TDQS
Scored across 14 tools
Most tools have distinct purposes (init, propose, proposal_decide, waivers, impact, explain), but the verification cluster—verify, check, and audit—overlaps significantly, with descriptions distinguishing them only subtly (build gate vs architecture rules vs full audit). Similarly, next_action, review, and analyze all produce or refresh bounded agent-facing context, creating some boundary confusion.
All 14 tools share a uniform lowercase snake_case `codapult_guard_` prefix, which is highly predictable. Slight inconsistency arises from mixing bare verbs (init, verify, propose, audit, check, review, explain, analyze) with noun forms (context, governance, waivers, impact, next_action, proposal_decide), but the pattern remains readable.
14 tools sits comfortably in the well-scoped 3-15 range and maps to discernible workflow stages (discover, init, propose, decide, verify, audit, review, manage waivers). No tool appears redundant enough to trim, and none of the surface feels padded.
The lifecycle is largely covered: discovery (context), initialization (init), rule proposal and approval (propose/proposal_decide), enforcement (verify/check/audit), review (review), exceptions (waivers), and introspection (impact, explain, governance, analyze). Minor gaps exist—no explicit reset/teardown of Guard artifacts and waiver management is a single bundled tool rather than discrete list/create/remove operations.