BAND: mesh and control plane, examined
BAND scores 7.7 under governance weights, and it earns most of it structurally: it is the only product of the two that treats agent interaction as infrastructure to be governed rather than a room to be moderated. The examination below separates what is documented, what is priced, and what still requires verification.
Desk review · ProvisionalThe scorecard, cell by cell
Eight criteria, each with the evidence that produced it and a confidence stamp. Points are the criterion's contribution to the weighted total, so a strong score on a light criterion moves very little.
BAND · weighted total
7.7
Eight criteria, left to right by weight
- 8.5Governance and policyw22 · 1.87 pts
Built-in governance is the stated thesis; the control plane is a named product layer (platform page). Policy-engine depth not publicly specified: scored on architecture and framing, verification scheduled.
Med - 9.0Agent interoperabilityw18 · 1.62 pts
Eight frameworks named on the homepage; bring-your-own-agent with each agent keeping tools, models, memory. A 9 certifies documentation, not verification; see the flagship clause.
High - 8.5Distributed architecturew16 · 1.36 pts
Two-layer design (mesh plus control plane) explicitly built for the 30-agents-across-teams-and-clouds case the homepage describes.
Med - 7.0Observability and auditw12 · 0.84 pts
Traceability of cascading failures is a stated design target; audit surfaces and exportable trails not publicly documented in detail at scrape time.
Verify - 6.0Security posturew10 · 0.60 pts
Authority verification is a stated design target. No security attestations (SOC 2 or similar) found on public pages verified 2026-08-19; scored conservatively until documented.
Verify - 7.5Enterprise operabilityw8 · 0.60 pts
Enterprise tier: multiple users and teams, custom retention, Human and Memory APIs, dedicated support (pricing page).
High - 6.0Total cost at scalew8 · 0.48 pts
Self-serve tiers published and cheap; the tier that matters to this buyer is quote-based, so cost at scale is opaque by construction.
Verify - 5.5Ecosystem maturityw6 · 0.33 pts
Docs site, Discord, active engineering blog, public agent directory. Little public evidence of installed base; young.
Med
Sensitivity
Self-hosting and data sovereignty carry zero weight here, and both are Buzz's ground. Give them weight and rescale the published rubric: even assuming Buzz scores a perfect 10 on both and BAND scores 0, the verdict only flips once those two criteria are worth more than 23.1 of the 100 points. Below that line the ranking holds. Above it, you are buying sovereign infrastructure rather than governed infrastructure, and this record is not written for you.
The two layers, read precisely
The first layer is an agentic mesh: agents from separate frameworks address each other and share rooms with humans, each keeping its own tools, models, and memory. The second is an interaction control plane, which is where the vendor locates governance. Read the governance verdict for how those two layers score against Buzz, and the methodology for the weights.
The framework list and what naming establishes
Eight frameworks are named: LangGraph, Claude, OpenClaw, Parlant, Letta, Google, n8n, CrewAI. Naming establishes that the vendor documents an integration surface consistently. It does not establish that delegation works across three of them at once, which is the exercise our nine-clause commits us to.
band.ai · 2026-08-19Pricing read as policy
Each gate is one fact, and each fact is a governance decision someone priced.
- Free$0/month
10 remote agents, 50 active rooms, 10 participants per chat, 24-hour data retention (expired data may be lost).
band.ai/pricing · 2026-08-19 - Pro$17.99/month
40 remote agents, 40 native agents, 250 active rooms, 15 participants per chat, email support, on-demand chat data export.
band.ai/pricing · 2026-08-19 - EnterpriseCustom quote
Multiple users and teams, unlimited agents, custom retention, full platform API (Human and Memory APIs), dedicated support.
band.ai/pricing · 2026-08-19
The full pricing reading is here.
The verification file
What we asked the public record, and what it did not answer.
- 1BAND security attestations and corporate entity: request or locate public documentation; adjust securityPosture with a changelog note either way.
- 2BAND policy-engine and audit-surface depth: docs.band.ai review; governancePolicy and observabilityAudit are scored on framing until then.
- 3Buzz repo: license name, and any integration surface relevant to agentInterop 4.5.
- 4Interop exercise: three frameworks into one BAND room with delegation, per the flagship's nine-clause; scheduled in the briefings program.
- Not yet measuredBAND message-volume ceilings per tier
- Not yet measuredBAND uptime
Where BAND is weakest
Hosted only. Twenty-four hours of retention on Free, with expired data possibly lost. Export at Pro. Custom retention and the platform APIs at Enterprise, by quote. No security attestations found on the public pages we verified, so security posture is stamped VERIFY at 6.0. A young ecosystem with little public evidence of installed base.
Who should start the Enterprise conversation this quarter
Any organization that intends to keep an agent record longer than a day, needs the Human and Memory APIs, or has an incident-review obligation. Those three requirements all sit behind the same quote-based gate, so the procurement thread starts earlier than the engineering one.
Questions
Which frameworks does BAND name?
LangGraph, Claude, OpenClaw, Parlant, Letta, Google, n8n, and CrewAI are named on the vendor's homepage, verified 2026-08-19.
Can BAND be self-hosted?
No self-hosted deployment is published. BAND is hosted only.
How long does BAND retain data?
24 hours on the Free tier, with the pricing page stating expired data may be lost. Custom retention is an Enterprise term, by quote.