Summary
If you're building agents that need to remember things across conversations, this gives you 16 tools for persistent context on Upstash Redis. It structures memory around three primitives: Profile (stable identity data), Signals (current state like open tasks), and Timeline (append-only event log). Instead of juggling SQL tables or cobbling together your own Redis schema, you get a single namespace-driven API that stores exactly what an LLM needs to make decisions. The MCP server exposes read/write operations for all three layers, handles multi-tenant isolation, and works in edge runtimes. Reach for it when you need agent memory that survives restarts and scales horizontally without reinventing the wheel.
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
Tools
Public tool metadata for what this MCP can expose to an agent.
17 tools
acmi_profileCreate or update an entity profile in ACMI. Stores arbitrary JSON profile data for an entity (agent, thread, project, etc.).3 params
Create or update an entity profile in ACMI. Stores arbitrary JSON profile data for an entity (agent, thread, project, etc.).
Parameters* required
idstringEntity ID within the namespace
profilestringJSON string of profile data to store
namespacestringACMI namespace (e.g. 'agent', 'thread', 'sales')
acmi_signalUpdate AI signals for an entity. Signals are mutable KV state (mood, priorities, scores, etc.) that changes frequently.3 params
Update AI signals for an entity. Signals are mutable KV state (mood, priorities, scores, etc.) that changes frequently.
Parameters* required
idstringEntity ID
signalsstringJSON string of signal data to store
namespacestringACMI namespace
acmi_eventLog a timeline event for an entity. The workhorse tool — records timestamped events with source, kind, correlationId, and summary. Follows ACMI Communication Standard v1.1.6 params
Log a timeline event for an entity. The workhorse tool — records timestamped events with source, kind, correlationId, and summary. Follows ACMI Communication Standard v1.1.
Parameters* required
idstringEntity ID
kindstringEvent kind (e.g. 'handoff-complete', 'step-done', 'decision')
sourcestringSource of the event (agent name, system, etc.)
summarystringHuman-readable event summary
namespacestringACMI namespace (e.g. 'thread', 'agent', 'work')
correlationIdstringCorrelation ID for tracking across agents/sessions (camelCase)
acmi_getFetch complete entity context: profile + signals + recent timeline events (last 10).2 params
Fetch complete entity context: profile + signals + recent timeline events (last 10).
Parameters* required
idstringEntity ID
namespacestringACMI namespace
acmi_listList all entity IDs in a namespace.1 params
List all entity IDs in a namespace.
Parameters* required
namespacestringACMI namespace to list
acmi_work_createCreate a new work item (cross-session project, task, or idea).2 params
Create a new work item (cross-session project, task, or idea).
Parameters* required
idstringUnique work item ID
profilestringJSON string of work item profile (title, owner, status, etc.)
acmi_work_eventLog a progress event on a work item.4 params
Log a progress event on a work item.
Parameters* required
idstringWork item ID
sourcestringSource of the event
summarystringEvent summary
sessionIdstringOptional session ID to associate
acmi_work_signalUpdate signals for a work item (progress, blockers, metrics, etc.).2 params
Update signals for a work item (progress, blockers, metrics, etc.).
Parameters* required
idstringWork item ID
signalsstringJSON string of signal data
acmi_work_getRead a work item's full context: profile, signals, timeline (last 50), and sessions.1 params
Read a work item's full context: profile, signals, timeline (last 50), and sessions.
Parameters* required
idstringWork item ID
acmi_work_listList all work item IDs.
List all work item IDs.
No parameter schema in public metadata yet.
acmi_catMulti-stream event merge view. Combines timeline events from multiple entities, sorted by timestamp. Supports --since filtering.3 params
Multi-stream event merge view. Combines timeline events from multiple entities, sorted by timestamp. Supports --since filtering.
Parameters* required
keysarrayTimeline keys to merge. Use 'thread:name', 'agent:name', or full 'acmi:...:timeline' keys.
limitnumberMax events to return. Default: 50.
sincestringTime window filter (e.g. '24h', '7d', '30m'). Default: all time.
acmi_spawnLog an agent session spawn event. Records when an agent starts a new session.3 params
Log an agent session spawn event. Records when an agent starts a new session.
Parameters* required
agentIdstringAgent ID that spawned
modelIdstringModel ID used for the session
sessionIdstringSession ID
acmi_bootstrapOne-shot agent context bundle. Fetches everything a fresh agent session needs: profile, signals, active threads, rollup, recent timeline, and spawns.1 params
One-shot agent context bundle. Fetches everything a fresh agent session needs: profile, signals, active threads, rollup, recent timeline, and spawns.
Parameters* required
agentIdstringAgent ID to bootstrap
acmi_activeTrack agent thread engagement. Add/remove threads or list current active threads for an agent.4 params
Track agent thread engagement. Add/remove threads or list current active threads for an agent.
Parameters* required
rolestringRole in thread (e.g. 'participant', 'lead'). Default: 'participant'
actionstringAction: add a thread, remove a thread, or list all active threadsone of
add · remove · listagentIdstringAgent ID
threadKeystringThread key (required for add/remove)
acmi_rollup_setSet the latest rollup snapshot for an agent (acmi:agent:<id>:rollup:latest). Pairs with acmi_bootstrap which reads it.2 params
Set the latest rollup snapshot for an agent (acmi:agent:<id>:rollup:latest). Pairs with acmi_bootstrap which reads it.
Parameters* required
rollupstringJSON string of rollup data (cross-session summary, decisions, blockers, etc.)
agentIdstringAgent ID
acmi_deleteDelete an ACMI key. Refuses protected paths (acmi:registry:*, acmi:notion-sync:*) and any non-acmi:* key. Defaults to dry-run; pass confirm=true to actually delete.2 params
Delete an ACMI key. Refuses protected paths (acmi:registry:*, acmi:notion-sync:*) and any non-acmi:* key. Defaults to dry-run; pass confirm=true to actually delete.
Parameters* required
keystringFull ACMI key to delete (must start with 'acmi:')
confirmbooleanMust be true to actually delete; otherwise returns dry-run preview. Default: false (dry-run).
acmi_search_semanticPerform semantic search across fleet coordination history. Finds relevant past events, decisions, and work items based on natural language queries. Returns original ACMI correlationIds for linking.2 params
Perform semantic search across fleet coordination history. Finds relevant past events, decisions, and work items based on natural language queries. Returns original ACMI correlationIds for linking.
Parameters* required
limitnumberNumber of results to return. Default: 5.
querystringNatural language search query (e.g. 'previous decisions about SSE timeouts')
ACMI - Agentic Context Memory Interface
The coordination backbone for AI agent fleets. Three Redis keys - Profile, Signals, Timeline.
ACMI is the open protocol for persistent agent context. Version v1.5 formalizes Fleet Comms Protocol: atomic pre/post events, wake-directives, handoff-ack chains, and correlation-aware timelines that make multi-agent work auditable instead of anecdotal.
Every entity stores exactly three things an LLM needs to make decisions:
Profile -> who (identity, preferences, stable facts)
Signals -> now (current state, blockers, next action)
Timeline -> then (append-only event log from every source)
The shape is intentionally small:
- Profile: stable identity and configuration.
- Signals: mutable state and synthesized working memory.
- Timeline: immutable history, ordered by time.
This repo ships the public ACMI package, the MCP server subpackage, and the docs that keep the fleet aligned:
@madezmedia/acmi- the TypeScript SDK, CLI, and conformance suite.mcp/-@madezmedia/acmi-mcp, the MCP server for hosts that need direct ACMI access.SPEC.md- canonical protocol spec.CHANGELOG.md- release history, including thev1.5.0fleet-comms update.docs/- operator guide, cheatsheet, and protocol notes.
What v1.5 adds
The v1.5.0 release aligns the fleet around a shared event language:
- atomic commit pre/post events
- roundtable coordination and wake-directives
source,kind,correlationId,summaryevent envelope discipline- signal freshness checks before action
agent:<id>source naming across the fleet
Install
npm install @madezmedia/acmi
Quick start
import { createAcmi } from "@madezmedia/acmi";
import { InMemoryAdapter } from "@madezmedia/acmi/adapters/in-memory";
const acmi = createAcmi(new InMemoryAdapter());
await acmi.profile.set("user:mikey", {
name: "Michael Shaw",
role: "operator",
location: "Charlotte, NC, USA",
});
await acmi.signals.set("user:mikey", "current_focus", "ACMI v1.5 fleet sync");
await acmi.timeline.append("user:mikey", {
ts: Date.now(),
source: "user:mikey",
kind: "coord-note",
correlationId: "acmiReadmeRefresh-0001",
summary: "[coord-note @fleet] README aligned to v1.5 and local assets.",
});
Production adapters
import { createAcmi } from "@madezmedia/acmi";
import { UpstashAdapter } from "@madezmedia/acmi/adapters/upstash";
const acmi = createAcmi(
new UpstashAdapter({
url: process.env.UPSTASH_REDIS_REST_URL!,
token: process.env.UPSTASH_REDIS_REST_TOKEN!,
})
);
| Adapter | Use case | Edge-compatible |
|---|---|---|
@madezmedia/acmi/adapters/in-memory | tests, examples, local dev | n/a |
@madezmedia/acmi/adapters/upstash | Vercel, Workers, edge runtimes | yes |
@madezmedia/acmi/adapters/redis | self-hosted Redis / Node | no |
Fleet protocol
ACMI v1.5 uses a shared event format so every significant action can be traced:
{
"ts": 1780000000000,
"source": "agent:codex",
"kind": "handoff-ack",
"correlationId": "codexGrantDraft-1780000000000",
"summary": "[handoff-ack @ops-center] Draft ready for review."
}
Rules that matter in practice:
- use
[kind-tag @recipient]in summaries - keep
sourceprefixed withagent:,user:, orsystem: - link follow-up events with
parentCorrelationId - keep the timeline append-only
- verify signals before acting when the workflow depends on current state
Related surfaces
- ACMI Product / live demo
- ACMI Operator Surface
- ACMI MCP server README
- Operator Guide
- ACMI Cheatsheet
- Changelog
The fleet
ACMI is used across the Mad EZ Media fleet as the common context layer for:
ops-center- orchestration and routingbentley- comms and governancecodex- coding and implementation supporthermes- deep scans and guardian checksandroid-worker- mobile bridge and notifications
License
MIT - Copyright Michael Shaw / Mad EZ Media
Featured
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →Configuration
UPSTASH_REDIS_REST_URL*Upstash Redis REST URL (https://your-instance.upstash.io)
UPSTASH_REDIS_REST_TOKEN*secretUpstash Redis REST token (read-write)
Categories
Registryactive
Package@madezmedia/acmi-mcp
TransportSTDIO, HTTP
AuthRequired
UpdatedMay 14, 2026
