Pragmabase is in public beta — Try the playground free →
← EnterpriseMulti-Agent Teams
Multi-Agent Teams

Agents that
collaborate,
not just coexist.

Shared memory. Pooled budgets. Role-based policies. Pragmabase teams are not just isolated agents running in parallel — they are coordinated systems with genuine shared context and economic accountability.

Team capabilities

Shared memory namespaces

Agents in a team share a Qdrant namespace. Knowledge written by one agent is immediately retrievable by any other. The team's understanding compounds over time.

Pooled + individual budgets

Define a team-level XLM budget and per-agent spending caps. Soroban guardrails enforce both. Agents can't overspend — individually or collectively.

Role-based agent policies

Assign roles to agents — coordinator, executor, auditor. Each role has distinct memory access permissions and budget authorities, enforced by Soroban.

Cross-agent task handoff

Agents pass context and partial work to teammates with full fidelity. The receiving agent picks up exactly where the previous one left off, memory and all.

Team audit trails

Every inter-agent communication, memory write, and payment is logged. Replay any decision tree across the entire team for debugging or compliance.

Parallel execution

Spawn parallel agent branches for time-sensitive workloads. Results are reconciled back into the shared memory namespace automatically.

How it works
01

Team provisioned

A team namespace is created with a shared Qdrant index, pooled Stellar wallet, and Soroban budget policy.

02

Agents spawned

Individual agents are assigned roles. Each gets a memory view and budget authority derived from the team policy.

03

Parallel execution

Agents run concurrently. Memory writes sync to the shared namespace in real time.

04

Budget enforced

Each x402 payment is validated against the agent's individual cap and the team's global budget by Soroban.

05

Audit captured

All decisions, payments, and memory mutations are logged immutably for replay and compliance.

// team: research-alpha

team.budget → 50 XLM
team.memory.namespace → "research-alpha"
 
agents:
gatherer → budget: 10 XLM, role: executor
analyst → budget: 20 XLM, role: executor
auditor → budget: 0 XLM, role: auditor
 
> gatherer.run("Fetch ETH/XLM data")
✓ wrote 3 memories to research-alpha
> analyst.run("Synthesize findings")
✓ retrieved 3 memories from team ns
✓ task complete (12.4 XLM team budget used)
Use cases

Research & synthesis teams

One agent gathers data, another synthesizes, a third writes the report — all sharing context and paying only for what they use.

Financial monitoring fleets

A coordinator agent dispatches market watchers. Any agent's finding is instantly available to the full team via shared memory.

Customer service pipelines

Triage agents classify requests and hand off to specialists, with full conversation memory transferred at handoff.

Audit & compliance agents

Dedicated auditor agents with read-only memory access monitor team activity and flag policy violations in real time.

Ready to run your first team?

Multi-agent teams are available on the Team plan and above. Enterprise customers get custom role policies and private memory isolation.