MULTI-AGENT COCKPIT / SHARED MEMORY
Your agents shouldn’t
start from zero.
Run Claude, Codex, local models, and shell agents side by side. Solomon carries the useful context between them—what changed, what failed, and what remains.
◆ Private early access. No noise, just launch updates.
FIG. 01 KNOWLEDGE MOVES. WORK CONTINUES.
› Trace why session restore fails after daemon restart.
Reading packages/daemon/src/resume.ts...
LEARNED The SDK error is misleading—the cwd is missing, not the binary.
Recording durable insight for the next agent.
One agent learns.
Every agent remembers.
WHY MEMORY
Insight shouldn’t die
when the session does.
Every agent run turns up hard-won context—the misleading error, the file that actually mattered, the fix that worked. Without shared memory, that context dies with the session. Solomon keeps it alive for whichever agent picks up next.
An agent works the problem and hits the real insight—not just the surface-level error.
Solomon records the durable takeaway, not the whole transcript—compact, reusable context.
The next agent—Claude, Codex, or a local model—reads it first and picks up mid-thread.
HOW IT WORKS
Four steps.
Zero re-discovery.
- 01Agent works the task
Claude, Codex, or a local model investigates, edits, and runs commands as usual.
- 02Writes a durable memory
The useful takeaway—what changed, what failed, what’s still open—gets recorded.
- 03Next agent reads first
Before starting, the next agent checks shared memory instead of re-tracing old ground.
- 04Project stays coherent
Every model works from the same evolving picture of the project, no matter who’s driving.
ENGINEERING RIGOR
Built for developers,
by developers.
No prototype energy. Here’s how we think about running Solomon somewhere your team can actually trust it.
Solomon is built to run against your own Cloudflare and model-provider accounts, so your project memory and agent activity stay inside infrastructure you already control.
Bring your own least-privilege API keys and connect only the scopes Solomon needs. Your code and prompts are never used to train models.
Built by engineers who run Claude, Codex, and local models against real codebases every day. Solomon is the tool we needed for our own multi-agent work, not a demo built for a pitch.
SSO, audit logs, and team-level access controls are on the roadmap as Solomon moves beyond private beta.
EARLY ACCESS
Give your agents
a memory.
Private beta. We’re onboarding a small number of teams running multi-agent workflows first.
◆ Private early access. No noise, just launch updates.