This project implements access to the v1 Meraki API
See the contribution guide for more information regarding contributing to this project.
This package's version tracks the Cisco Meraki Dashboard API version it targets. It is not semantic versioning.
1.70.79
│ │ └── build height, incremented automatically by Nerdbank.GitVersioning
│ └───── Meraki Dashboard API minor version (this package targets API v1.70)
└──────── Meraki Dashboard API major version
So 1.70.x means "generated against Meraki Dashboard API v1.70". The major and minor parts move
when this library is regenerated against a newer Meraki API, and only then.
The practical consequence: do not bump the major version to signal a breaking change in this
library. Releasing a 2.0.x would announce a Meraki Dashboard API v2.0 that does not exist.
Breaking changes to this library are called out in CHANGELOG.md instead, and the
version continues to follow Meraki.
When updating to a new Meraki API version, see the guidance in .github/copilot-instructions.md.
net10.0 only. Panoramic Data does not support anything lower.
The netstandard2.0 target was dropped after 1.70.79. It existed to serve .NET Framework and older
.NET consumers, but it also forced a second, quietly different build: several methods could not pass
a CancellationToken, and the retry logic compared error strings case-sensitively on
netstandard2.0 and case-insensitively on net10.0, so the two targets could disagree about
whether to retry. The MCP client was excluded from it entirely. Consumers who need netstandard2.0
should stay on 1.70.79.
Full documentation is available here including:
- how to use the library
- all methods and properties
MerakiClient.Workflows is a typed client for the complete
Cisco Workflows Automation API.
It covers all 156 published operations and their request and response models, including workflows,
runs, targets, variables, schedules, rules, triggers, tables, and Exchange import/export.
using var meraki = new MerakiClient(new MerakiClientOptions
{
ApiKey = merakiDashboardApiKey,
UserAgent = "YourProduct/1.0 YourCompany"
});
var page = await meraki.Workflows.GetAllWorkflowsPostQueryParamsAsync(
organizationId,
limit: 100);
foreach (var workflow in page.Results ?? [])
{
Console.WriteLine($"{workflow.Id}: {workflow.Title}");
}Every operation takes organizationId because Workflows routes are organization-scoped. The
existing MerakiClientOptions region, read-only guard, rate limiter, retry/back-off, logging, and
statistics apply unchanged. API keys are automatically sent as Authorization: Bearer, as required
by Workflows, without changing authentication for Dashboard API calls.
See the Workflows API guide for listing, creating, validating, executing, inspecting, and cleaning up a minimal workflow, plus multipart upload, prerequisites, versioning, and the opt-in live integration test.
In addition to the REST client, this package can act as a client for the Cisco Meraki MCP server, letting an agent you build search the Meraki Dashboard API capability catalogue in natural language and execute the capability it selects.
using Meraki.Api.Mcp;
// The Cisco-hosted server (the default).
await using var client = new MerakiMcpClient(new MerakiMcpClientOptions
{
ApiKey = merakiDashboardApiKey
});
var capabilities = await client.SemanticSearchAsync("which clients are on this network?");
var result = await client.ExecuteApiAsync(
capabilities[0].CapabilityId,
new Dictionary<string, object?> { ["networkId"] = "N_123" });
// The server wraps results in an envelope. Payload is the "data" element alone, which
// matches the shape the equivalent MerakiClient call returns. Deserialize<T>() uses it.
Console.WriteLine(result.Payload);
var clients = result.Deserialize<List<Client>>();A self-hosted instance of the open-source server over HTTP:
await using var client = new MerakiMcpClient(new MerakiMcpClientOptions
{
Transport = MerakiMcpTransport.LocalHttp,
Uri = "http://localhost:8080/mcp",
ApiKey = merakiDashboardApiKey,
ApiRegion = ApiRegion.Government
});Or launched as a child process over stdio:
var options = new MerakiMcpClientOptions
{
Transport = MerakiMcpTransport.Stdio,
Command = "npx",
ApiKey = merakiDashboardApiKey
};
options.Arguments.Add("cisco-meraki-mcp-official");
await using var client = new MerakiMcpClient(options);Use GetStatusAsync() for a non-destructive connectivity and credential check.
Two things observed against the live hosted server that are worth knowing:
MerakiCapability.Scoreis currently alwaysnull— the beta server returns no relevance score, so the ranking is the list order alone. Do not filter on a score threshold.- The server reports some failures, such as a capability called without its required parameters, in
the payload of an otherwise successful tool result. Both operations detect that and throw
MerakiMcpProtocolExceptionincluding the server's recovery suggestion, so if a call returns, you have data rather than an error document.
These come from Cisco's own documentation, and are worth reading before you design around the server:
- The server is in beta, and Cisco documents that breaking changes are possible.
- Operations are read-only.
execute_apicannot mutate anything. UseMerakiClientfor writes;ExecuteApiAsyncrefuses non-read capabilities rather than letting you discover this at runtime. - Authentication is by API key only. OAuth is not yet supported.
- The hosted server supports Meraki.com environments only — not Federal, GovCloud, or localised
instances. Those require a self-hosted server, so
MerakiMcpTransport.HostedHttprejects a non-defaultApiRegion. - Six static egress IP addresses may require allowlisting for the hosted server. They are listed
on
MerakiMcpTransportException.HostedEgressIpAddresses, and named in the exception message when a connection times out. - Rate limits are shared. The server respects the Dashboard limit of 10 requests per second per
organization, and that budget is shared with any
MerakiClienttraffic in the same process.
Meraki's Dashboard API rate limit is per organization, and it is shared with every other
consumer of that organization's API: other applications, scripts, and monitoring tools such as Splunk
that poll aggressively. When the limit is hit the API answers 429 Too Many Requests with a
Retry-After header.
MerakiClient retries 429s (and 502/503/504s, per-attempt timeouts and transient connection
failures) transparently. The relevant MerakiClientOptions are:
| Option | Default | Meaning |
|---|---|---|
MaxAttemptCount |
500 | Attempts per call before giving up |
BackOffDelayFactor |
1.0 | Base of the exponential back-off. 1.0 is a flat delay |
MaxBackOffDelaySeconds |
30 | Ceiling on any single wait |
HttpClientInnerTimeoutSeconds |
25 | Timeout for one attempt |
HttpClientTimeoutSeconds |
600 | Timeout for the whole call, spanning all attempts and waits |
Before attempt n the client waits min(max(BackOffDelayFactor^(n-1), Retry-After), MaxBackOffDelaySeconds)
seconds plus a small upward jitter. At the default factor of 1.0 that is simply Retry-After, or one
second where the server sends none.
The defaults are deliberately persistent. Against a saturated organization, growing back-off is
counter-productive: the longer this client waits, the more of the shared budget the other consumers
take. The most reliable way to get a call through is to keep re-asking, honouring Retry-After,
until a slot opens. So out of the box a call most likely completes, at the cost of possibly taking a
long time. In practice the bound on that time is HttpClientTimeoutSeconds, which ends a still-throttled
call with a TaskCanceledException long before attempt 500 is reached.
To fail fast instead, override the defaults:
var client = new MerakiClient(new MerakiClientOptions
{
ApiKey = apiKey,
UserAgent = "YourProduct/1.0 YourCompany",
MaxAttemptCount = 5, // give up after five attempts
BackOffDelayFactor = 2.0, // 1s, 2s, 4s, 8s between them (capped by MaxBackOffDelaySeconds)
HttpClientTimeoutSeconds = 60 // and never spend more than a minute on one call
});When attempts run out the client returns the last response, so Refit raises an ordinary ApiException
carrying that status (typically 429). That response is indistinguishable from a first-attempt 429; if
you would rather be told "throttled for ten minutes" than "HTTP 429", set
ThrowOnRetryExhaustion = true and the client throws RetryExhaustedException instead, carrying the
attempt count, the time spent and the final status. It is off by default and is not an ApiException,
so existing catch (ApiException) handlers will not see it. Per-attempt timeouts that run out of
attempts throw TimeoutException.
Endpoints that return large collections are paginated by the Meraki API using startingAfter /
endingBefore cursors and a Link: rel=next header. Each such endpoint has a matching *AllAsync
extension method (for example GetOrganizationNetworksAllAsync) that follows the links and returns
the whole collection as one list.
These helpers are all-or-nothing. If any page fails, the exception propagates and every page already fetched is discarded, so an exception means "no data was returned", never "here is what was fetched so far". Two consequences for calling code:
- A 404 from a genuinely empty collection and a 404 on the last of many pages reach the caller as the
same
ApiException. Do not treat an exception from an*AllAsynccall as an empty result. Code that reconciles a local cache against the returned list should retry or abort on exception rather than delete everything. - If the API advertises a next page the client cannot follow, the helpers throw
PaginationExceptionrather than return a silently truncated list.
Per-page methods (the same name without All) return one page and leave cursor handling to you.
The Meraki API documentation can be found here:
You can test this using a Meraki Sandbox here:
After signing in, look in the lower left hand side of the page for your API key.