How Partners Sell LinkMesh

If you integrate infrastructure, run observability for other people, or get called in when a monitoring bill arrives, LinkMesh is straightforward to sell — but not as a product pitch. It sells as the answer to a problem your clients describe to you before they have a name for it: telemetry volume grows faster than the budget, and nobody can say what next quarter’s invoice will be.

This is the partner-facing version of that argument: who to bring it to, what to say in the first meeting, the commercial numbers you can quote without checking, and the four objections you will actually hear.

The one problem it solves

Observability backends are priced per gigabyte. That means the cost of watching a system scales with the system, not with the value of what is being watched — and the decision that sets the volume (what gets collected and shipped) happens at the edge, weeks before the invoice that reflects it.

LinkMesh is a self-hosted control plane for OpenTelemetry Collectors. It manages the collectors in front of the backends: filter what the client does not need, mask what must not leave the host, route the rest to wherever it belongs — and it is priced per Collector rather than per gigabyte, so volume never moves the bill.

That is the whole pitch, and it is worth saying in that order: problem, what it is, what changes. Everything below is detail underneath it.

Who to qualify — and who to walk away from

The signals that a client is a fit, in rough order of how strongly they predict one:

• They run their own infrastructure, because they are regulated, air-gapped, or simply unwilling to ship raw telemetry to a vendor.
• Twenty to two hundred collectors or monitoring agents. Below that the free tier serves them and there is no deal; above it they are still a fit, and the bill stops growing (see below).
• More than one backend. Splunk for the SOC, Sentinel because it came with E5, Grafana because the platform team likes it. That estate is normal and it is permanent — see one telemetry strategy, multiple backends.
• An agent per vendor on every host. Four agents reading the same files is the slide that gets your project approved.
• An auditor or a data protection officer asking questions about what leaves the network and who reviewed it.
• A renewal or a budget conversation in the next two quarters. This is the trigger; the rest is qualification.

Who is not a fit: a client buying hosted observability SaaS and happy with it, and anyone with a handful of hosts — the free tier covers them, and quoting a deal there wastes your meeting and their goodwill.

The three-step pitch

The plan is three steps, and it is the same three everywhere, because a client who can repeat it back can sell it internally without you in the room:

1. Install the control plane — one command on a Linux VM inside the client’s own infrastructure.
2. Point the collectors at it — one config block; the standard OpenTelemetry Collector over OpAMP. No proprietary agent to adopt, no fork.
3. Filter, mask and route — cut what they do not need, mask what must not leave the host, send the rest where it belongs.

What you can promise in return, all three defensible: it is self-hosted, so their telemetry never reaches us; it is priced per Collector, so volume never changes the bill; it is standard OpenTelemetry, so they can leave.

For the first meeting, step one is the demo. It takes minutes — the quickstart is the whole of it — and a client watching their own collector appear in a fleet view is a better argument than any slide about control planes.

The numbers you can quote

All of this is public on the pricing page, so you can say it in a meeting and the client can verify it on their phone:

• USD 150 / CHF 144 / EUR 150 per Collector per year — that is USD 12.50 a month, billed annually.
• The first 25 Collectors are free, deducted from every invoice, not just the first. A free Community licence is issued on registration in the customer portal; an unregistered install runs five Collectors without a licence at all.
• The bill stops growing at 200 Collectors — a flat USD 26,250 / CHF 25,200 / EUR 26,250 per year, with unlimited Collectors above that. 250, 2,500 or 25,000 cost the same as 200.
• The cap is early-access pricing and it has an end date. Early access ends on 30 November 2027 and the cap changes with it. A subscription started before then keeps the 200-Collector cap for as long as it renews.

That last point is the only genuine piece of urgency in the pitch, and it is real rather than manufactured — which is exactly why it is worth using. A client with a large fleet and a 2027 budget cycle has a reason to start the evaluation now.

For a client coming off a volume-priced contract, the Splunk cost calculator turns their own numbers into the comparison, which is a far better opening than a price list.

Where your margin actually is

Be clear with yourself about this, because it shapes which deals are worth your time: there is no published channel programme, margin schedule or deal-registration portal. If you want to resell rather than refer, that is a conversation per arrangement — start it through contact.

What is reliably yours is the work around the licence, and on a fleet of this size it is the larger number:

• The assessment. Counting agents per host, mapping what is collected to what is actually queried, and putting a figure on the gap. This is billable on its own and it is what produces the business case.
• Migrating off proprietary agents. Retiring a vendor agent per source, keeping the collection layer constant — see migrating Splunk Enterprise to OpenTelemetry for the shape of that project.
• Pipeline design. Parsing, enrichment, redaction rules and routing per destination. This is where domain knowledge of the client’s estate is worth more than product knowledge of LinkMesh.
• Running it. A control plane someone has to operate, upgrade and answer for is a managed service, and an obvious annuity next to a one-off integration.

If you operate fleets for several clients, one install can carry more than one estate: collectors live in groups, roles can be scoped to a single group, and routing is per-tenant — the architecture choices are covered in multi-tenant Collector architecture. Whether you run one shared install or one per client is a commercial decision as much as a technical one, and worth talking through before you quote it.

The four objections, and the honest answers

• “Is this another agent we have to adopt?” No. It drives standard OpenTelemetry Collectors over OpAMP — no proprietary agent, no fork. If they already run otelcol-contrib, they already run the data plane; see onboarding otelcol over OpAMP.
• “Can you see our telemetry?” No. It is self-hosted by design: we have no access to their telemetry, their configs or their dashboards. For a regulated client this is usually the sentence that ends the security conversation rather than starting it.
• “Are you certified?” Not yet, and our trust page says so in plain language — no SOC 2 and no ISO 27001 today. Do not improvise around this one. For this audience, a vendor that publishes what it does not have is more credible than a badge wall, and the compatibility page publishes the product’s limitations for the same reason. LinkMesh is built by OpenSight, an independent Swiss company — a named team, which is the thing the certification question is usually really asking.
• “What if we want to leave?” The collectors are standard and the config is yours. Nothing proprietary sits in the telemetry path, so backends stay swappable and so does the control plane. The honest version of this answer is what makes the first three believable.

How to run the first meeting

1. Ask what they ingest per day and what they pay per gigabyte. Most clients know one of the two. The gap between those numbers is the conversation.
2. Count the agents on one representative host with them, in the meeting.
3. Install the control plane on a VM and enrol one collector. Minutes, in their infrastructure, with nothing leaving it.
4. Filter one genuinely noisy source and show the before and after. A rule previewed against a real captured sample, not a promise about one.
5. Then put the numbers on it — their volume, the per-Collector price, the cap. In that order, because the price only lands once the saving is theirs rather than hypothetical.

If a client wants to shape where the product goes rather than just buy it, the design partner programme is the route, and it is a stronger offer than a discount for the kind of engineering-led client you want as a reference.