Incident management · Production

LinkedIn's Contextual Agent Playbooks & Tools (CAPT) gives AI coding agents organizational context for 1,000+ engineers

The problem

AI coding agents lacked the organizational context needed to help LinkedIn engineers with company-specific systems, frameworks, and workflows, limiting them to generic coding tasks and leaving engineers hesitant to rely on them for real work.

First attempt

Existing AI coding assistants like GitHub Copilot handled generic tasks but broke down on LinkedIn-specific workflows, and an initial namespace-based approach to organizing CAPT tools still pushed complexity onto engineers who had to manually manage which namespaces to enable.

Workflow diagram · grounded in source
1
Engineer submits natural language request
Trigger
Engineers, PMs, and EMs start with a natural language question to initiate the workflow.
▾ source quote
“engineers, PMs, and EMs can start with a natural language question, have an agent translate it into the right queries against our data platforms”
2
LLM discovers and routes to appropriate tools
Routing
Meta-tools let the LLM discover tools by tag, inspect their schemas, and execute them.
▾ source quote
“These meta-tools let the LLM (not the user) discover tools by tag, inspect their schemas, and execute them (get_tools_for_tags, get_tool_info, exec_tool). Each underlying tool is tagged by function (e.g. experimentation, logs, metrics, deployments) and the LLM uses those tags to …”
3
Playbook orchestrates multi-system investigation
Ai action
The agent queries metrics, logs, deployment history, and incident records through CAPT's tools.
▾ source quote
“the agent queries metrics, logs, deployment history, and incident records, looks for recent rollouts or related failures”
4
Narrative surfaced to engineer
Output
The agent surfaces a narrative covering what changed, what is breaking, and where to look first.
▾ source quote
“surfaces a narrative: what changed, what is breaking, and where to look first”
5
Human reviewer approves final output
Human review
Final approval remains with the human reviewer.
▾ source quote
“while still leaving the final approval to a human reviewer”
Reported outcome

CAPT is used by more than 1,000 LinkedIn engineers; issue triage time dropped by about 70%, data analysis is roughly three times faster, and over 500 playbooks have been authored company-wide.

Reported metrics
engineers using CAPTmore than 1,000
Issue triage time reduction (summary)about 70%
Initial customer issue triage time reductionaround 70%
Data analysis speed improvementroughly three times faster
Show all 9 reported metrics
engineers using CAPTmore than 1,000
issue triage time reduction (summary)about 70%
initial customer issue triage time reductionaround 70%
data analysis speed improvementroughly three times faster
time from question to usable insightroughly 3× faster
playbooks authored company-wideOver 500
data pipeline and ML job debugging timemore than half
analysis turnaroundanalysis that used to take days of back-and-forth can now be done in hours
engineer weekly time savings from code reviewseveral hours a week
Reported stack
Model Context Protocol (MCP)GitHub CopilotJinja2Trino
◆ Does this fit your context?

Compare to your context

Tell us your scale, team, and constraints. We'll show what changes at your size, what fails at your scale, and whether this case is a fit, needs adaptation, or won't scale to you. Free demo, no signup.

Compare to your context →
~30 seconds · free
Source
https://www.linkedin.com/blog/engineering/ai/contextual-agent-playbooks-and-tools-how-linkedin-gave-ai-coding-agents-organizational-context
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

CAPT is used by more than 1,000 LinkedIn engineers; issue triage time dropped by about 70%, data analysis is roughly three times faster, and over 500 playbooks have been authored company-wide.

What tools did this team use?

Model Context Protocol (MCP), GitHub Copilot, Jinja2, Trino.

What results were reported?

engineers using CAPT: more than 1,000; Issue triage time reduction (summary): about 70%; Initial customer issue triage time reduction: around 70%; Data analysis speed improvement: roughly three times faster (source-reported, not independently verified).

What failed first in this deployment?

Existing AI coding assistants like GitHub Copilot handled generic tasks but broke down on LinkedIn-specific workflows, and an initial namespace-based approach to organizing CAPT tools still pushed complexity onto engi…

How is this incident management AI workflow structured?

Engineer submits natural language request → LLM discovers and routes to appropriate tools → Playbook orchestrates multi-system investigation → Narrative surfaced to engineer → Human reviewer approves final output.

Related incident management cases

WHAT TO DO WITH THIS

Now compare it to your context

This case is one data point. Whether its pattern fits you depends on your volumes, your stack, and your exception load — that comparison is the step no case study can do for you.

Browse the Incident management cases →