Skip to main content
Glama

Synapse Layer — Trust Infrastructure for AI Agents

by io.github.SynapseLayer

recall

Read-only

Retrieve relevant persisted memory using semantic, temporal, priority, or hybrid routing. Governance: requires reason (10–200 chars). Rate limit: 20/min.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
modeNoRecall routing mode. "temporal" = pure chronological, "semantic" = FTS + TQ ranking, "priority" = critical/high-TQ first, "hybrid" = weighted blend, "auto" = detect from query (default).
limitNoMaximum memories to return (1–50, default: 10).
queryYesWhat to recall — natural language query for memory retrieval.
scopeNoExplicit memory scope. "agent" (default, fail-closed) = limits recall to the given agent_id (or the token agent). "tenant" = cross-agent recall within the authenticated tenant. If omitted with agent_id present, "agent" is inferred; if both omitted, the token agent is used (verified-token scope).
reasonYesREQUIRED (10–200 chars). Audit policy: MEMORY_EXPORT_GOVERNANCE_V1. Human-readable justification for retrieving PLAINTEXT memory. Without this parameter the call WILL fail with code -32602. Example: "User asked me to summarize last week decisions."
subjectNoESCOPO1 (optional): opaque 8-128 char token ([A-Za-z0-9._:-]) that isolates memory to a single person within a shared connect-token. Omit for token-scoped behavior. Never include PII/email.
agent_idNoAgent identifier (canonical or alias). Used only when scope="agent" or scope is inferred.

Output Schema

TableJSON Schema
NameRequiredDescriptionDefault
countYes
scopeYes
memoriesYes

TDQS

A3.5/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare readOnlyHint=true and openWorldHint=false. The description adds valuable behavioral context beyond this: the mandatory 'reason' parameter with character range, the governance/audit policy implication, and the 20/min rate limit. This is meaningful additional disclosure of operational constraints.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Two concise sentences with no filler. The core action is front-loaded, followed immediately by the critical governance constraint and rate limit. Every phrase earns its place and the structure is easily scannable.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

With an output schema present and 100% schema coverage, the return format and parameter semantics are already documented. However, the description lacks context about the fail-closed scope behavior and, more importantly, does not disambiguate from the 'recall_memory' sibling, leaving a notable completeness gap for an AI agent selecting between tools.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, and the schema already provides rich semantics for each parameter, including enums and examples. The description's mention of 'reason (10–200 chars)' reinforces the schema but does not add significant new meaning beyond what the parameter descriptions already state. Baseline 3 is appropriate.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb ('Retrieve') and resource ('persisted memory') along with routing modes, which clearly identifies the tool's function. However, the sibling tool 'recall_memory' exists with a nearly identical name, and the description does not differentiate this tool from it, so it is not a 5.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description explains governance and rate limits but gives no guidance on when to use 'recall' versus the closely related 'recall_memory', 'search', or 'list_memories' siblings. There is no mention of conditions, exclusions, or alternative routing, leaving the agent to infer usage from the name and schema.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All Synapse Layer — Trust Infrastructure for AI Agents toolsInspect
Try in Browser
RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

TDQS

B3.1/5.0
Disambiguation2/5

Several tools are redundant: recall_memory is explicitly an alias of recall, and save_memory/save_to_synapse/store_memory all persist memories. This makes the tool boundary unclear for agents choosing between them, even though core operations like health_check and slo_report are distinct.

Naming Consistency2/5

Names are consistently snake_case, but verb usage is inconsistent: save_memory, save_to_synapse, and store_memory use different verbs for the same operation, and recall/recall_memory add an alias rather than following one convention. The pattern is readable but not predictable.

Tool Count3/5

Thirteen tools is not inherently excessive for an agent-memory/trust service, but the presence of multiple aliases inflates the surface and means not every tool earns its place. The effective set is smaller, making the count feel padded.

Completeness3/5

The surface covers initialization, save, list, search/recall, feedback, and handover, but there is no update or delete operation for memories. This leaves lifecycle management incomplete and may force agents to work around missing state-management operations.

Resources