Frontier
Memory portability between AI providers
Move your compiled memory between models or vendors — provenance and validity windows travel with it.
96+ ideas to build
Build with durable memory
Statewave is an open-source memory runtime for AI agents. Support agents are the strongest workflow today — but the same primitives power coding copilots, account assistants, voice continuity, multi-agent platforms, and far more. This page is the map.
The unifying loop
Every use case on this page is the same loop
Existing systems become episodes. Episodes compile into typed memory. Memory gets ranked into a token-bounded bundle. Your agent gets the context it needs — and nothing it doesn’t.
01
Existing data
Tickets, code, docs, calls, events
02
Episodes
Immutable, append-only
03
Compiled memory
Typed facts with provenance
04
Context bundles
Ranked, token-bounded
Strongest today
The proven wedge: support agents
Support is where Statewave is deeply optimized — session-aware ranking, handoff packs, health and SLA scoring, and repeat-issue detection. It’s the workflow with the most evidence today, not the limit of the platform.
Support & serviceStrongest today
Support agent memory
Persistent customer context across sessions — plan, history, prior issues, last resolution. The flagship Statewave workflow.
Support & serviceStrongest today
Escalation handoff packs
Compact, structured briefs with active issue, attempted steps, health, and SLA — ready for tier-2 or human takeover.
Support & serviceStrongest today
Support health & SLA intelligence
Deterministic 0–100 health scores with explainable factors. First-response and resolution SLA tracking with breach detection.
Support & serviceStrongest today
Repeat-issue detection
Surface prior resolutions automatically when recurring patterns appear. Stop re-solving the same ticket.
Use case explorer
Beyond support: what else fits
The same memory primitives power developer copilots, workspace assistants, account intelligence, voice continuity, and multi-agent infrastructure. Filter by category or by how mature each shape is today.
Category
Status
Showing 4 of 54 ideas · Available
Connectors & bootstrap
Statewave is not just for live chats
You don’t need to wait for a new conversation to start. Pipe existing system history into episodes, compile it once, and your agent walks into its first session already informed.
Official ecosystem
Looking for ready-made connectors instead of integration recipes? Statewave Connectors are modular packages — install only what you need. Phase 1 ships GitHub, Markdown/docs, and the MCP server.
The bootstrap pattern
Existing data
ticket / repo / doc / call
Episodes
immutable, per-subject
Compiled memory
typed, confidence-scored
Better context
ranked, token-bounded
Every connector below is a recipe for that pattern. Most teams start with a one-shot historical import, then keep their connector running incrementally so the memory stays current as new events arrive.
View Statewave ConnectorsStatus
Support systems
Backfill ticket history into episodes — bootstrap memory before the first live session.
Zendesk ticket import
Backfill historical tickets and comments as episodes per customer (organization or requester) subject.
Intercom conversation import
Ingest conversations, public replies, and admin notes scoped per customer (primary company or contact). US/EU/AU regions.
Freshdesk ticket import
Backfill historical tickets, public replies, and private notes scoped per customer (company or requester). API key auth.
Engineering systems
Ingest issues, PRs, and commits to give coding agents real project memory.
GitHub issues + PRs
Compile decisions, review comments, and resolutions as durable repo memory.
Docs & wikis
Compile durable knowledge from internal documentation rather than re-embedding it ad hoc.
Notion page ingestion
Treat Notion pages (and optionally their body content) as decision-memory episodes; re-pull when last_edited_time advances. Subject is operator-controlled — repo:owner/name, project:foo, or any string.
GitHub markdown docs
Sync /docs and README content as episodes so the agent always speaks the latest API.
CRM & customer notes
Bring account history into a structured per-account memory.
Slack channel + thread import
Targeted channel imports — internal conversations as searchable episodes per subject.
Gmail relationship memory
Pull messages matching a Gmail search query — per-contact email history scoped to relationship:<email>. Body extracted from MIME (text/plain preferred, text/html fallback). OAuth 2.0 refresh-token; gmail.readonly scope.
Cards tagged Recipe are integration patterns you build on Statewave’s ingest API — write a small importer in your preferred language; the SDKs make the per-subject episode loop straightforward. Cards tagged Available or Coming soon ship as official packages — see the .
Frontier ideas
Ideas at the edge
Directions Statewave makes possible but doesn’t yet ship pre-built. Some are research territory; some are just engineering work the platform doesn’t prescribe. Treat this section as inspiration, not a roadmap.
Frontier
Audit-grade memory for regulated workflows
Provenance-traced facts and immutable episodes as the substrate for compliance-grade AI applications.
Frontier
Federated memory across orgs
Cross-organization memory protocols where each side keeps storage local but agreed-upon facts can be shared.
Frontier
Personal memory layer for end-users
Consumer agents that remember you across apps — under your control, deletable, and transparent.
Frontier
Industry-specific compiled memory packs
Pre-trained compilers and memory schemas for verticals — legal, healthcare, finance, support.
Frontier
Agent-to-agent memory protocols
A standard handshake for two agents to exchange typed memory, not unstructured prose.
Frontier
Memory-aware evals across vendors
Evaluate models on the same compiled memory layer — apples-to-apples context quality benchmarks.
Frontier
Continuous self-improving copilots
Copilots that compile their own mistakes into preference memories and quietly get better each week.
FAQ
Frequently asked questions
What is Statewave strongest at today?
Support agents with returning customers. That workflow has the most machinery behind it: session-aware ranking, escalation handoff packs, repeat-issue detection, first-response and resolution SLA tracking, and deterministic 0–100 health scores with explainable factors (healthy ≥70, watch 40–69, at_risk below 40). It is the shape with the most evidence today, not the limit of the runtime — the same primitives carry every other use case on this page.
Can I use Statewave outside customer support?
Yes. The explorer on this page catalogs 54 use-case ideas across coding copilots, workspace and account assistants, voice continuity, and multi-agent infrastructure, four of them written up as full deep-dives. They differ only in which subjects you write and which task you retrieve for; record, compile, retrieve, and govern is the same loop underneath all of them.
Does my agent have to start from an empty memory?
No. The bootstrap pattern backfills the history you already have: pipe existing tickets, repositories, documents, and call transcripts through the connectors as episodes, compile once, and the agent walks into its first session already informed. Because episodes are append-only and compilation is idempotent, a backfill can be re-run safely — it produces no duplicate memories.
Which use case should I build first?
Pick one subject type and one workflow — a customer, a repo, or a pipeline run — and wrap the two-call pattern around the agent you already have: retrieve a context bundle before the model call, write an episode after it. That is enough to see ranked memory in production. Widen to more subjects and more connectors once the first loop is boring, rather than modelling every subject up front.
Answers last checked against the Statewave docs and repositories on .
EXPLORE THE POSSIBILITIES
See your idea on this page
If you’re building any AI workflow with multi-session memory, Statewave is the layer underneath. Run it locally in two minutes — or try the live demo first.