Stateful agents.
Zero rewrites.

Trove is a proxy for agent memory. It learns from every call, injects what's relevant into the next, and forwards to your provider. One line to add it.

client = OpenAI(base_url="https://api.openai.com/v1") # Traffic through Trove builds memory. Your provider API keys stay yours.
Works with OpenAI Anthropic LangChain CrewAI Vercel AI SDK

Your agents already produce the memory.

It resets every time a session ends.

i.

Your code review agent flagged the deprecated auth pattern. Your bug triage agent rediscovered it three days later. That's not an edge case. That's every sprint.

ii.

Agents have no shared awareness. What Agent A resolves disappears the moment the session ends. Agent B starts from zero, every time.

iii.

Existing memory tools mean an SDK install, schema design, custom plumbing, and ops work. Weeks before the first useful memory.

iv.

When an agent makes a wrong call, there's no record of what it knew, or didn't know, at that moment. Debugging is archaeology, unless you can replay what was in the window.

How it works.

Invisible to your agents. Auditable in the logs. No rewrites.

Every request passes through Trove.

No agent code changes. Point your client at the Trove gateway. Memory builds from every prompt and response on that path; Trove forwards the call to your provider with your API keys.

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.trove.dev/v1"
)

Cross-agent context

Agent B opened with the right hypothesis.

Agent A flagged the auth issue on Monday. Trove carried it to triage on Thursday.

code-review-agent Mon 09:12
trove stored context
bug-triage-agent Thu 14:08

Flagged deprecated auth in api/session.ts. Team decision: migrate to JWT middleware.

Scroll to play

Not a database. A context engine.

Storage is a commodity. What matters is what context surfaces, and when.

Context assembly engine

On every request, Trove chooses what goes in the window, not your team. Recency- and relevance-weighted, packed to your token budget for this agent, this task, this moment.

Automatic entity extraction

Users, tickets, services, decisions parsed from every gateway call. No manual tagging.

Structured knowledge graph

Derived from the event log: typed relationships with timestamps. This is what Trove searches when it enriches the next call.

Cross-agent awareness

What one agent learns is available to the next. No shared-memory contracts or handoff glue.

Append-only event log

Raw record of every gateway call. Replay and audit exactly what was prepended when a decision shipped.

Provider-neutral

OpenAI today, Anthropic tomorrow. Switch models without losing state or history.

Every competitor is building a feature.

Trove is neutral infrastructure. The one position a model provider can't take.

Trove OpenAI / Anthropic Vector DBs LangChain / CrewAI Memory SDKs
Multi-agent, cross-session memory ~ ~ ~
Provider-neutral infrastructure ~ ~
One-line proxy integration
Context assembly (not just storage) ~ ~ ~
Framework-agnostic ~
Stateful agents in under 5 minutes

Fully supported ~ Partial / limited Not available

One gateway. Any provider.

Same base_url swap for OpenAI, Anthropic, or LangChain. Or self-host the proxy in Docker. Trove holds the memory layer; your provider keys stay yours.

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.trove.dev/v1"
)

Built with the first teams who hit this wall.

We're working directly with a small group of engineering teams running agents in production. Early partners shape the architecture, not just the roadmap.

No sales call. No pitch deck. We'll reach out when the beta opens.