Skip to main content
Connic

Agent observability.
Built into Connic.

Follow model calls, tools, and MCP calls through each run. Connic records timings, errors, and token usage, with costs and operational metrics in the dashboard.

Start for freeRead the guide
Read the observability docs

Run details

#a1b2c3d4
completed
2.34s1,234 tokens$0.018Dec 5 · 14:34:12
  1. Runinvoice-processor2,341msok
  2. LLMconnic/gpt-5.6-sol · 1,203 tokens1,892msok
  3. Tooldocuments.parse234msok
  4. MCPlinear.create_issue156msok
  5. Tooldatabase.store_invoice59msok

From the full run to individual calls

The trace shows the steps of an agent run and their timings. Select a step to inspect its inputs and outputs. The summary shows token usage and total cost.

Run details

#a1b2c3d4
completed
2.34s1,234 tokens$0.018Apr 12 · 14:34:12
  1. Runinvoice-processor2,341msok
  2. LLMconnic/gpt-5.6-sol · 1,203 tokens1,892msok
  3. Tooldocuments.parse234msok
  4. MCPlinear.create_issue656msok
  5. Tooldatabase.store_invoice59msok
LLM · connic/gpt-5.6-sol · 1,203 tokens
ok
input
system_prompt + invoice text + tool schemas
output
tool_call: documents.parse(s3://invoices/...)
followed by tool_call: database.store_invoice(...)

First LLM hop. Token cost dominates the run.

Keep track of ongoing operations

Connic brings together success rate, run counts, token usage, and average costs. See when costs rise, failures become more frequent, or usage changes.

Success rate

98.5% · selected range

Average token cost per run

$0.018 · across all runs

Agent runs

completed vs. failed

Usage by model

gpt-5.6-sol / claude-sonnet-5 tokens

Narrow down failures and investigate causes

Connic helps find affected runs, spot unusually high costs, and compare failed executions with successful ones.

Isolate the failing slice

In the Logs tab, filter runs by status, deployment, date range, and context data, such as customer IDs or request types added by middleware.

Logs search: customer_id=abc123
Catch runaway-cost runs

Anomaly detection compares each run against the agent's 30-day rolling cost average and sends a notification when one breaks out. Spend limits pause the agent if a daily or monthly threshold is crossed.

Anomaly: run cost > 3× rolling avg
Compare a bad run to a good one

A failing run and a healthy run can be opened side by side. The first divergent span often points to the underlying cause.

First-divergence: LLM, Tool, or MCP span

What Connic includes for observability

Which capabilities are built in, and which need to be added? Compare Connic with LangSmith, Helicone, and a custom OpenTelemetry setup.

What Connic includes for observability
FeatureConnicLangSmithHeliconeDIY OTel
Per-run trace treeIncludedIncludedPartialPartial
Cost per run / per agentIncludedIncludedIncludedNot included
Token breakdown (input / output / thinking / cached)IncludedPartialPartialNot included
Captured logs from tools & middlewareIncludedPartialNot includedPartial
Custom dashboards in-productIncludedPartialPartialPartial
Spend alerts & hard limitsIncludedNot includedPartialNot included
Cost anomaly detectionIncludedNot includedNot includedNot included
Re-run any execution from the dashboardIncludedPartialNot includedNot included
Wired to A/B testingIncludedNot includedNot includedNot included
Wired to judgesIncludedPartialNot includedNot included

Keep exploring

Judges

Score every run. Catch drift early.

A/B testing

Compare variants in production.

Approvals

Human-in-the-loop on sensitive calls.

Dev server

Hot-reload local agent runtime.

Testing

Assertions and judges, in CI.

Connectors

Wire agents to real systems.

Frequently Asked Questions

Connic captures spans for LLM calls, Tool invocations, MCP tool calls, Middleware before/after hooks, Triggered Agent links, and the top-level Run itself. Each sequential step is a separate linked run with its own trace. Each span shows status, duration, inputs, outputs, and metadata. LLM spans also include the model's reasoning whenever the provider returns it (controlled via reasoning_effort on the agent).

Every run records input, output, thinking, and cached-input tokens. connic/* charges are shown in EUR. BYOK usage keeps the currency configured for that model, so USD and EUR remain separate when both occur.

The Token Usage page supports currency-specific alerts and hard limits against token cost. Both can be scoped globally, per environment, or per agent, and use a daily or monthly period. Project Billing separately provides Project credit alerts and auto-refill controls.

Yes. When one agent triggers another, the child run's header shows a 'Triggered by' link back to the parent. The full chain is navigable from there.

The project Logs tab supports key=value search against run context data. Context data such as customer_id, request_type, or workflow_version can be added by middleware to find the corresponding runs.

Yes. Run Again in any run detail view re-triggers the same agent with the same input. Queued or running runs also have a Cancel button.

Yes. Middleware after() hooks receive a context dict with run_id, agent_name, duration_ms, and token_usage when the agent finishes. That data can be sent by POST to an external monitoring system. Output from print or Python's logging module in tools, middleware, hooks, and guardrails is also captured per run and surfaced in the Logs tab.